Research

INTRODUCTION

Roads form a central component of human society. They connect people and communities, facilitating access to employment, resource extraction, and trade (Meijer et al. 2018, Weiss et al. 2018). It is projected that by 2050, a 60% increase in combined road and rail networks is needed to meet expected increases in global mobility through trade and travel (Dulac 2013). Despite their obvious benefits, roads have many negative environmental impacts. They facilitate human colonization (Ibisch et al. 2016), cause wildlife road mortality, fragment habitats, create barrier effects, and reduce habitat quality (Huijser and Bergers 2000, Jaeger et al. 2005, Seiler 2005, Jacobson et al. 2016, Abra et al. 2019, Teixeira et al. 2020). Ultimately, these impacts result in loss of biodiversity and affect ecosystems negatively (Ibisch et al. 2016).

Studying the effects of roads and other linear infrastructure on wildlife has become a major focus of ecological research, particularly concerning population- and community-level effects (Fahrig and Rytwinski 2009, van der Ree et al. 2011). These effects are influenced by species-specific behavioral responses to roads and traffic volume, categorized by Jacobson et al. (2016) into four groups: “speeders,” “pausers,” “avoiders,” and “non-responders.” For example, as “speeders,” mule deer (Odocoileus hemonius) have evolved to flee from danger, exploiting gaps in traffic; thus, their likelihood of road mortality increases with increased traffic volume until a threshold is reached at which there is a barrier to movement (Jacobson et al. 2016). Importantly, wildlife road mortalities involving larger bodied “speeders” also pose a severe risk to the traveling public (Jaeger et al. 2005, Huijser et al. 2009, Ford et al. 2022). Consequently, collisions with common ungulates (ungulate–vehicle collisons; UVCs) such as deer are typically framed as public safety issues rather than conservation concerns (Seiler and Helldin 2006). In Canada, an estimated 45,000 reported collisions with large-bodied wildlife occur annually, resulting in 30 to 40 human fatalities each year (Meister et al. 2016). In the province of Alberta alone, wildlife–vehicle collisions account for 43% of collisions on rural roads (Government of Alberta 2020b). These collisions carry significant direct and indirect costs for society, including medical expenses, human injuries or fatalities, vehicle repairs, emergency personnel and police attendance, lost hunting revenue, and carcass disposal expenses (Huijser et al. 2009).

Because the occurrence of UVCs on roads raises both ecological and public safety concerns, it highlights the need for a robust data collection framework to assess collision trends in space and time (Schwartz et al. 2020). Since 2009, there has been substantial advancement in wildlife–vehicle collision data collection and analysis systems at large scales suitable for transportation agencies (Shilling and Waetjen 2015, Bíl et al. 2017, 2020a, Shilling et al. 2020). Decision support tools are also available, such as cost-benefit analyses specifically designed to evaluate wildlife mitigation projects (Huijser et al. 2009, Wilkins et al. 2019). Despite these advances in understanding animal–vehicle collision reduction, challenges persist because of the short-term and often geographically limited nature of mitigation studies, which lack inferential strength (Roedenbeck et al. 2007, Ford et al. 2022). Generalizations about where collisions occur have been further complicated by the recognition that major biophysical gradients play a more significant role than previously thought (Clevenger et al. 2015). Additionally, conflicts arise among stakeholders, including government agencies, nongovernmental organizations, and the public, as they struggle with how to prioritize limited resources in addressing areas of concern. The social and geographic contexts of roads also play vital roles, affecting actor interdependency and history of collaboration, and the complexity of governance in each location (Huber 2022).

Identifying ungulate–vehicle collision hotspots

UVC hotspots are neither spatially nor temporally random (Seiler 2005, Hussain et al. 2007, Ng et al. 2008, Gunson et al. 2009, 2011, Chen and Wu 2014). At large scales, ungulate density and traffic volume are important predictors of interannual trends in UVCs (Lavsund and Sandegren 1991, Groot Bruinderink and Hazebroek 1996, Seiler 2004, Rolandsen et al. 2011, Bíl et al. 2021a, Shilling et al. 2021). Other factors that influence UVC rates include local vegetation and food availability (Keken et al. 2019), landscape characteristics (Seiler 2005, Gunson et al. 2011), road density (Ng et al. 2008), road type and ditch or verge topography (Hubbard et al. 2000, Seiler 2005), environmental conditions (e.g., weather; Huijser et al. 2008), proximity to human development (Finder et al. 1999), wildlife population dynamics (Groot Bruinderink and Hazebroek 1996), and driver response (Vanlaar et al. 2019). Despite these documented predictors, some interannual temporal and spatial variability in UVCs is expected due to random effects or unmeasured variables both at and beyond the road (Santos et al. 2016).

Several analytical approaches are used to identify collision hotspots based on road mortality or traffic crash data. The most appropriate procedure may vary depending on the management objective (Clevenger et al. 2006). Suitable statistical methods include Getis Ord Gi* and spatial autocorrelation (Shilling and Waetjen 2015), incident density (Shilling and Waetjen 2015, Bíl et al. 2017), kernel density estimation (KDE; Anderson 2009), KDE-plus (KDE+; Bíl et al. 2013), Dangerousness Index (Steenberghen et al. 2010), Nearest Neighbour Index (Biggs et al. 2004), and others (Clevenger et al. 2006). In addition, there are tools (software) available within which statistical methods have been already implemented, for example, Crimestat (Levine 2010), KDE+ (Bíl et al. 2016), and Siriema (Coelho et al. 2014).

Here, we evaluate the spatiotemporal evolution of UVC hotspots and changes to UVC road mortality in response to major road construction in two study areas. First, we apply cluster analysis to detect UVC hotspots (Bíl et al. 2013, 2016). Our approach is novel in that we also test the spatiotemporal KDE+ (STKDE+) analytic tool (Bíl et al. 2019a) to detect stable UVC collision hotspots over space and time and to support decision-making. STKDE+ applies the KDE+ method in a user-selected moving temporal window and examines whether a hotspot has newly developed, disappeared, or remained stable for the study period (Fig. 1). Whereas other clustering methods are applied to a certain time interval (e.g., 4 yr), the STKDE+ can be applied to the entire period of the data; the STKDE+ output then indicates where and when problems (or a successful measure) took place. We also take into account changes in traffic volume and regional wildlife population trends. We expected that some UVC hotspots would be (1) temporally stable, whereas others would (2) emerge, disappear, and re-emerge over time, particularly as part of larger global zones of collisions, or (3) change in response to human development, wildlife mitigation measures, wildlife population dynamics, or changes to road alignment.

DATA AND METHODS

Case study 1: new highway construction with wildlife mitigation

The first study area was located directly south of the City of Calgary, Alberta, Canada in the Parkland Natural Region and included a pair of highway segments: an unfenced highway (MacLeod Trail or Highway 2A) and a fenced bypass (Deerfoot Trail or Highway 2) linked at their south end by a Y-intersection (Fig. 2). A new section of the fenced bypass opened in December 2003 as part of a major highway project that bypassed the unfenced highway (Carter 2005). The fenced bypass has a posted speed limit of 110 km/h and included the construction of approximately 4.5 km of wildlife fencing associated with two wildlife crossing structures: (1) a river bridge with 30 m clearance for wildlife passage on either bank, and (2) an additional 85 m long steel-plate wildlife underpass (7 m wide by 4 m high) with a 2 m diameter skylight (Carter 2005; Fig. 3). Both highways pass through parkland habitat with a mix of urban and acreage development; however, the fenced bypass descends and crosses the Bow River, whereas the unfenced highway follows the crest of the southwest side of the river valley. In the southeastern part of Wildlife Management Unit 212, which includes this study area, there are locally abundant populations of white-tailed deer (O. virginianus) and mule deer (O. hemonius), and a stable population of elk (Cervus canadensis; Government of Alberta 2020a). Moose (Alces alces) populations have increased in this area, related to a range expansion into the Parkland Natural Region of Alberta that started in the 1980s (Bjorge et al. 2018).

Case study 2: twinning of an existing highway

The second study area was located on Highway 63, which acts as the main transport route to Fort McMurray in northern Alberta. In response to public safety concerns about the high frequency of motor vehicle accidents along this corridor (almost half of which involved wildlife), approximately 240 km of the highway was converted from two-lane undivided to four-lane divided highway. Construction began with widening the median in 2006, with the most disruption to normal traffic use occurring from 2012–2015 (Fig. 4). The highway expansion included a 250-m right-of-way and an increase in the posted speed limit from 100 to 110 km/h. We focused on 54.5 km of the construction zone because ungulate population surveys (specifically, white-tailed deer and moose), traffic volume, and UVC data were available for our time interval of interest.

Although agricultural land dominates the southern end of the segment, the highway also bisects a variety of boreal mixed wood and lakeland habitat types (Donker and Chapman 2015). Changes to the landscape through resource extraction activities have facilitated the invasion of white-tailed deer into the boreal forest (Fisher et al. 2020), such as are found in proximity to the northern part of the study highway. In response, wolf populations have increased, at the expense of woodland caribou populations (Rangifer tarandus caribou; Latham et al. 2011). Regional wildlife population surveys in Wildlife Management Unit 503 estimated a stable white-tailed deer population with a density of 1.62 deer/km² (Donker and Chapman 2015). The moose population was estimated to be 0.3 moose/km² and increasing (Donker and Chapman 2015). Both mule deer and elk are also present in the area at low densities. Woodland caribou, a species at risk, is present at very low densities (Donker and Chapman 2015). No associated wildlife fencing or underpasses were installed during highway construction. We were interested in examining the effect of the new construction (divided highway with an increase in posted speed limit) on spatial UVC patterns.

Data sources

Government-managed animal road mortality data were accessed from (1) police collision reports provided by Alberta Transportation and (2) animal carcass reports available through open access at https://www.alberta.ca/open-government-program. We used police collision data from 2000–2018 because it was available with equal sampling effort across the province and its records include geospatial data since 2000. The Government of Alberta requires motorists to report any collision to police if the accident results in a minimum cost threshold of ≥ $1000 CAD (before 2011; $774 USD) and ≥ $2000 CAD (since 2011; $1548 USD). Because of these repair thresholds, we expected that police collision databases under-report animal–vehicle collisions (Clevenger et al. 2003) and that they are biased toward collisions involving larger bodied animals (National Academies of Sciences, Engineering and Medicine 2007). Thus, we treated this analysis as an index of trend rather than representing total actual collisions.

Despite their aforementioned limitations, long-term and large-extent police collision databases can be used for larger scale analysis of UVC hotspots (Clevenger et al. 2006, 2009) because hotspots are spatially non-random (Snow et al. 2015, Bíl et al. 2017). The species or family of animal (e.g., deer) involved was parsed from the comments section, where provided. We removed all non-ungulates, including “unknown.” Spatial coordinates were provided for each report; however, we recognized that the reported location for police collisions is often referenced to a landmark or a road kilometre mark. In a study of the spatial accuracy of roadmark-based wildlife–vehicle collision data vs. landmark-referenced data, Gunson et al. (2009) reported that the mean spatial error of wildlife–vehicle collision locations using the closest landmark was 516 m (standard deviation [SD], 808 m; range, 0–6500 m), whereas a road marker data set had a mean distance of 401 m (SD, 219 m; range, 7–794 m). Thus, we assumed an average UVC spatial accuracy of ± 500 m and incorporated a ± 500-m buffer into spatiotemporal cluster analyses.

Where animal carcass data were available at similar reporting levels to historical animal–vehicle collision data, we extended the spatiotemporal analysis to 2021. We used the following attributes from this presence-only data: precise location, date, and species. We did not use animal carcass data for comparing the day of week or time of day for collisions because carcass pick-up may not occur on the same day as an accident takes place. We also expected that animal carcass records underreport carcasses because some animals who are hit move away from the roadside before dying (Lee et al. 2021).

Data analysis

For data visualization, we used KDE+ software (Bíl et al. 2016, 2021b) as an objective tool to identify spatially non-random UVC clusters in each study segment. The method involves separating the road network into between-intersection segments, relating each collision record to its location along a respective road segment, and then applying a statistical significance test to identify and rank hotspots objectively.

Bíl et al. (2019a) recently developed a spatiotemporal version of the KDE+ method known as STKDE+. This method applies KDE+ in a user-selected moving temporal window and examines whether a hotspot has newly developed, disappeared, or remained stable for the study period (Fig. 1). We applied the STKDE+ tool using a 3-yr sliding window on either side of each collision report along the length of each study site. We focused specifically on road mortalities and not the use of crossing structures. All calculations were carried out in the KDE+ and STKDE+ toolboxes for ArcGIS (https://www.kdeplus.cz/en/download). For all analyses, divided highways were treated as a single lane.

We used a before-after comparison for existing highways and a control-impact for new highway construction to evaluate changes to ungulate mortality rates. Equal time intervals and highway lengths were used for each before-after or control-impact test.

In case study 1, 13.071 km were assessed for UVC hotspots. For the control-impact (MacLeod Trail-unfenced vs. Deerfoot Trail-fenced bypass, respectively) analysis, we subset the data and compared 4.5 km of the fenced (impact) and control highways. We used three years (2004–2006) for the control-impact analysis because traffic volumes were similar for both control and impact sites at that time. The analysis was repeated adding UVCs within a 500-m buffer at each end of the wildlife fence at the impact site. For the control unfenced highway (MacLeod Trail), we also conducted a before-after comparison of UVC rates in response to the drop in traffic volume after the bypass opened. We corrected the UVC rates from the before data set using regional wildlife collision trends (within the Wildlife Management Unit) as an indicator of overall population changes.

For case study 2, a 54.5-km segment was assessed for UVC hotspots. For the before-after analysis, we compared UVC rates from 2009–2011 to those from 2015–2018 because traffic volumes were similar and government estimates of moose and white-tailed deer populations were available during both time intervals.

For both study areas, Shapiro-Wilk tests were applied to determine normality (null hypothesis: data are normally distributed using a significance level of α = 0.05). For normally distributed data, an independent samples t-test was used to compare before-after or control-impact UVC rates (measured as collisions km¹ yr¹; null hypothesis: no change using a significance level of α = 0.05). Ungulate population estimates were accessed from published government surveys or inferred by examining total annual UVC within the Wildlife Management Units encompassing each highway. Traffic counters were located within each study highway, and traffic data were accessed at https://www.alberta.ca/highway-traffic-counts. Traffic data were reported as average annual daily traffic volume (AADT). We used a Wilcoxon rank sums test to see if UVC rates differed on weekdays vs. weekends.

The raw effect size of each mitigation measure or road change on UVC frequency was calculated using the standardized Hedges’ g:

(1)

Where M₁ denotes the mean UVC km⁻¹; yr⁻¹ for the control site or before mitigation conditions; M₂ denotes the mean UVC km⁻¹; yr⁻¹ for the impact or after interval; and SD_pooled denotes the pooled standard deviation, which was calculated as:

(2)

Where n_i denotes the number of collisions and s_i² denotes the variance in the ith time interval. For samples with N ≤ 50, Hedges’ g was modified as follows to account for the upward bias of small sample size:

(3)

Where N denotes the total combined collisions.

RESULTS

A total of 930 UVCs were compiled over both study areas, 315 and 615 for case studies 1 and 2, respectively (see Appendix 1 for annual UVC totals). For the control-impact and before-after analyses, subsets of 35 UVCs were used for case study 1 and 235 UVCs for case study 2.

Case study 1: new highway construction with wildlife mitigation

After the fenced (impact, Deerfoot) bypass opened, the average AADT on the unfenced (control, MacLeod Trail) dropped from 32,623 vehicles (median = 32,910 vehicles; SD = 890 vehicles, 2000–2003) to 19,957 vehicles (median = 20,410 vehicles, SD = 1792 vehicles, Fig. 5). The fenced bypass became the preferred route for motorists, with its AADT increasing from 18,970 vehicles to a maximum of 45,560 vehicles in 2017 (Fig. 5).

For the control-impact comparison, there were significantly lower UVC rates (an 86% difference) on the fenced bypass (impact) vs. the unfenced (control) highway (t₄ = 5.59, P = 0.0025), with a large effect size (g = 3.3), even when UVCs within a 500-m buffer were included (t₄ = 5.14, P = 0.032, g = 3.2). The decrease in traffic volume on the unfenced highway after the bypass opened had a small effect (g = 0.1) on UVC rates, but it was not significant (t₆ = 0.47, P = 0.655). Because the traffic volume became extremely high (> 32,000 AADT) on the fenced bypass, the probability of a UVC became higher on weekends vs. weekdays (W = 34, P = 0.0096).

A series of UVC hotspots were identified in the study area (Fig. 6). Hotspot “A” (Fig. 6) on the unfenced highway was 1.39 km in length and it remained stable over time. When the fenced bypass opened in December 2003, a hotspot 0.480 km in length (see “B,” Fig. 6) developed at the south end, outside the exclusion fence. Its location is aligned with the southern edge of an acreage subdivision. Two additional hotspots were also identified within the exclusion fence area, which were in the vicinity of a wildlife corridor along the south slope (north-facing) of the river valley and in proximity to an access gate within the exclusion fence.

When smartphone-based carcass collection with photo identification to species level began in 2017 (N = 72), 60% of carcasses on the unfenced highway were mule deer, 34% were white-tailed deer, and 5% were moose. On the fenced bypass, 54% of carcass reports were mule deer, 46% were white-tailed deer, and there were no reports of moose–vehicle collisions.

Case study 2: twinning of an existing highway

There was no significant difference in the mean traffic volume in the before (M = 3860 vehicles, SD = 331 vehicles) vs. after (M = 4177 vehicles, SD = 111 vehicles) construction time intervals (t₄ = −1.25, P = 0.2794; Fig. 7). After construction was complete, the reemergence of historical UVC hotspots at several locations along the highway was evident (Fig. 8). A total of six UVC hotspots were identified, with 27.74% of UVCs occurring in hotspots, and hotspots comprising 6.69% of the segment length. Although UVCs increased by 9.3% post-construction (before M = 0.69 UVC km⁻¹; yr⁻¹, SD = 0.09 UVC km⁻¹; yr⁻¹ vs. after M = 0.75 UVC km⁻¹; yr⁻¹, SD = 0.01 UVC km⁻¹; yr⁻¹), the difference was not significant (t₂ = −0.50078, P = 0.642825), and the effect size was small (g = −0.2).

DISCUSSION

Our findings provide insight into various aspects of UVCs. The effectiveness of 4.5 km of wildlife fencing with two underpasses was evident in an 86% difference in UVCs compared to a control. However, at a larger scale, the proximity to an unfenced high traffic volume road resulted in a net increase in UVCs. We also found that increasing the number of lanes and posted speed limit on a highway resulted in a slight increase in UVCs, and that many preexisting UVC hotspots reemerged after construction was completed, when no mitigation measures for wildlife had been included. Our results demonstrate the utility of running cluster analyses through time to provide a deeper understanding of where UVC hotspots occur, how they change over time, and the impacts of wildlife mitigation and road realignment.

Wildlife fencing with associated underpasses has a significant effect on reducing direct wildlife mortality, with longer stretches of fence (≥ 5 km) having greater effectiveness than fences < 5 km (~80% vs. 52%) with less variability (Huijser et al. 2016). Our results support the notion that 4.5 km of wildlife exclusion fence with two underpasses in an urban area has a large effect on reducing direct ungulate mortality at the highway scale. However, while not an objective of this study, it is worth mentioning that the net benefits of the wildlife crossing structures and fencing of the bypass highway were negatively affected by road density, specifically by the proximity to an unmitigated high-volume highway. We found that the overall number of UVCs increased at a regional scale, which is consistent with other studies that report a positive correlation between road density and wildlife–vehicle collisions (Groot Bruinderink and Hazebroek 1996, Seiler 2005). Huijser and Begley (2022) stress the importance of choosing the appropriate scale at which to evaluate wildlife mitigation measures, especially in terms of net benefits.

We demonstrated that at AADT of 20,000 to 32,000 vehicles, direct road mortality is the major effect of highways on local deer populations, whereas at higher traffic volumes (AADT > 32,000 vehicles), deer mortality decreased, suggesting a barrier effect due to traffic as per Jacobson et al. (2016). We also provided empirical support for the notion that at extremely high traffic volumes, mule deer and white-tailed deer may experience higher mortality on days when traffic volumes and traffic intensity are lower during their normal peak active periods (e.g., dawn and dusk), such as on weekends in this case. In a review of white-tailed deer–vehicle collision literature, Steiner et al. (2014) note that the difference in deer–vehicle collision frequency on weekends vs. weekdays is related to differences in traffic use patterns and intensity during daily periods of peak deer activity, rather than biological, ecological, or behavioral factors. This trend has also been observed with roe deer (Capreolus capreolus, Bíl et al. 2020b, Ignatavičius et al. 2020).

We found that monitoring UVC trends in a variety of ways over space and time provides several advantages. Grouping by a predefined time interval is required for before-after-control-impact analyses. However, the addition of hotspot monitoring using STKDE+ provided an easy visualization of UVC hotspot locations, emergence, stability, and response to road construction. STKDE+ also highlighted the emergence and stability of a fence-end hotspot, where an interchange near the fence end separated ungulates from easy access to the nearest wildlife underpass. The emergence of UVC hotspots in proximity to fence ends has been reported for a variety of species in other studies (Clevenger et al. 2001, Plante et al. 2019, Huijser and Begley 2022).

Increasing the number of lanes and posted speed limit had a small negative effect on UVC rates, with slight, but not significant, increases in both deer and moose vehicle collisions. This result is consistent with what is reported elsewhere: in a review by Pagany (2020), which included factors influencing animal–vehicle collisions, 80% of applicable studies found a positive relationship between wildlife–vehicle collision risk and the number of lanes (which is a proxy for traffic volume).

For each study area, ungulate density (and by extension, UVC rates) will be affected by landscape-level factors (Clevenger et al. 2015). For example, deer frequently occur in high densities in proximity to urban areas because of attractive forage opportunities, often resulting in deer–vehicle collisions (Magle et al. 2014). Similarly, in the boreal forest, Fisher et al. (2020) found that anthropogenic landscape change best predicted seasonal white-tailed deer habitat selection. White-tailed deer strongly select for linear features in the boreal forest (Fuller et al. 2023) and throughout the year (Darlington et al. 2022). The large right-of-way associated with widening highways through the boreal and mixed-wood forests may also increase the availability of attractive roadside forage associated with early seral vegetation for white-tailed deer (Darlington et al. 2022), which may increase the potential for UVCs. Additionally, a decrease in winter severity due to climate change has reduced winter mortality in deer in the boreal forest that typically occurred due to the energetic stress associated with historically cold winters and associated deep snow (Dawe et al. 2014).

Hotspot analytical methods

Calculating stepwise spatiotemporal variation in UVC frequency can be computationally challenging (Bíl et al. 2020b) and difficult to incorporate into some UVC hotspot models (Visintin et al. 2016). Analysis of UVC clusters using programs such as KDE+ (Bíl et al. 2016) and STKDE+ (Bíl et al. 2019a) provide separate but complimentary views of UVC hotspots. These types of analysis can also be incorporated into dashboard-based reporting systems, which are increasingly becoming the norm in transportation and citizen science reporting (see https://albertawildlifewatch.ca/, https://www.kdebourame.cz/en/, and others; Shilling et al. 2020). For example, in Alberta Wildlife Watch (Shilling et al. 2020), a cluster analysis is automatically completed nightly, enabling managers to identify and respond to issues in a timely manner. Both KDE+ and STKDE+ can also be applied at a variety of scales, from individual highway segments to larger geographic areas (Bíl et al. 2019a,b).

Study limitations

We were constrained to using police collision data because they were available at the desired temporal scale for this study, and we recognize their limitations with underreporting. We also recognize that the lack of a full before-after-control-impact study design reduced the inferential strength of our before-after and control-impact comparisons; the main added value of our analysis was the use of STKDE+ for spatiotemporal UVC hotspot monitoring.

CONCLUSION

Our study highlights the importance of spatiotemporal scale in planning and evaluating wildlife mitigation measures on roads. We recognize that while road mortality data and UVC hotspot analyses are important, they do not necessarily indicate where wildlife crossing opportunities are most needed for priority species (Jacobson et al. 2016, Paemelaere et al. 2023). We recommend the use of planning frameworks that also incorporate the behavioral responses of priority species to roads and traffic volume (Jacobson et al. 2016) and species- or guild-specific ecological considerations (Paemelaere et al. 2023; Kintsch et al., unpublished manuscript: https://arc-solutions.org/wp-content/uploads/2021/03/08.-ICOET-WildlifeCrossingGuilds-paper.pdf). By monitoring UVC hotspot evolution over space and time, we were able to identify stable vs. ephemeral hotspots, detect a fence-end and barrier effect for ungulates, and explore hotspot stability before and after construction. Mitigation measures such as the 4.5 km of wildlife fencing and two underpasses on the mitigated impact highway in our study greatly reduced UVCs and increased driver safety in comparison to the unmitigated control. However, wildlife mitigation measures are also recommended on the nearby unmitigated highway. Furthermore, the conversion of a two-lane undivided road to a four-lane divided highway and simultaneous increase in the posted speed limit had limited impact on unmitigated spatiotemporally stable UVC hotspots. Instead, we suspect it led to increased driver speed, as evidenced by the slight increase in UVC rates.

LITERATURE CITED

Abra, F. D., B. M. Granziera, M. P. Huijser, K. M. P. M. B. Ferraz, C. M. Haddad, and R. M. Paolino. 2019. Pay or prevent? Human safety, costs to society and legal perspectives on animal-vehicle collisions in São Paulo state, Brazil. PLoS ONE 14(4):e0215152. https://doi.org/10.1371/journal.pone.0215152

Anderson, T. K. 2009. Kernel density estimation and K-means clustering to profile road accident hotspots. Accident Analysis and Prevention 41(3):359-364. https://doi.org/10.1016/j.aap.2008.12.014

Biggs, J., S. Sherwood, S. Michalak, L. Hansen, and C. Bare. 2004. Animal-related vehicle accidents at the Los Alamos National Laboratory, New Mexico. Southwestern Naturalist 49(3):384-394. https://doi.org/10.1894/0038-4909(2004)049<0384:AVAATL>2.0.CO;2

Bíl, M., R. Andrášik, V. Cícha, A. Arnon, M. Kruuse, J. Langbein, A. Náhlik, M. Niemi, B. Pokorny, V. J. Colino-Rabanal, C. M. Rolandsen, and A. Seiler. 2021a. COVID-19 related travel restrictions prevented numerous wildlife deaths on roads: a comparative analysis of results from 11 countries. Biological Conservation 256:109076. https://doi.org/10.1016/j.biocon.2021.109076

Bíl, M., R. Andrášik, M. Duľa, and J. Sedoník. 2019b. On reliable identification of factors influencing wildlife-vehicle collisions along roads. Journal of Environmental Management 237:297-304. https://doi.org/10.1016/j.jenvman.2019.02.076

Bíl, M., R. Andrášik, and Z. Janoška. 2013. Identification of hazardous road locations of traffic accidents by means of kernel density estimation and cluster significance evaluation. Accident Analysis and Prevention 55:265-273. https://doi.org/10.1016/j.aap.2013.03.003

Bíl, M., R. Andrášik, and J. Sedoník. 2019a. A detailed spatiotemporal analysis of traffic crash hotspots. Applied Geography 107:82-90. https://doi.org/10.1016/j.apgeog.2019.04.008

Bíl, M., R. Andrášik, and J. Sedoník. 2021b. STKDE+ method. Version 1.0. [Computer software]. Transport Research Centre, Olomouc, Czech Republic. http://kdeplus.cz/en/stmethod

Bíl, M., R. Andrášik, T. Svoboda, and J. Sedoník. 2016. The KDE+ software: a tool for effective identification and ranking of animal-vehicle collision hotspots along networks. Landscape Ecology 31(2):231-237. https://doi.org/10.1007/s10980-015-0265-6

Bíl, M., F. Heigl, Z. Janoška, D. Vercayie, and S. E. Perkins. 2020a. Benefits and challenges of collaborating with volunteers: examples from National Wildlife Roadkill Reporting Systems in Europe. Journal for Nature Conservation 54:125798. https://doi.org/10.1016/j.jnc.2020.125798

Bíl, M., J. Kubeček, and R. Andrášik. 2020b. Ungulate-vehicle collision risk and traffic volume on roads. European Journal of Wildlife Research 66:59. https://doi.org/10.1007/s10344-020-01397-8

Bíl, M., J. Kubeček, J. Sedoník, and R. Andrášik. 2017. Srazenazver.cz: a system for evidence of animal-vehicle collisions along transportation networks. Biological Conservation 213(A):167-174. https://doi.org/10.1016/j.biocon.2017.07.012

Bjorge, R. R., D. Anderson, E. Herdman, and S. Stevens. 2018. Status and management of moose in the parkland and grassland natural regions of Alberta. Alces: a Journal Devoted to the Biology and Management of Moose 54:71-84. https://www.alcesjournal.org/index.php/alces/article/view/232

Carter, D. 2005. Deerfoot Trail extension – innovative approach leads to geometric innovations. In Proceedings of the 2005 annual conference of the Transportation Association of Canada. Conference of the Transportation Association of Canada, Ottawa, Canada. http://conf.tac-atc.ca/english/resourcecentre/readingroom/conference/conf2005/docs/s11/carter.pdf

Chen, X., and S. Wu. 2014. Examining patterns of animal-vehicle collisions in Alabama, USA. Human-Wildlife Interactions 8(2):9. https://doi.org/10.26077/p18k-1089

Clevenger, A. P., M. Barreuto, K. E. Gunson, F. M. Caryl, and A. T. Ford. 2015. Context-dependent effects on spatial variation in deer-vehicle collisions. Ecosphere 6(4):47. http://dx.doi.org/10.1890/ES14-00228.1

Clevenger, A. P., B. Chruszcz, and K. E. Gunson. 2001. Highway mitigation fencing reduces wildlife-vehicle collisions. Wildlife Society Bulletin 29(2):646-653. https://www.jstor.org/stable/3784191

Clevenger, A. P., B. Chruszcz, and K. E. Gunson. 2003. Spatial patterns and factors influencing small vertebrate fauna road-kill aggregations. Biological Conservation 109(1):15-26. https://doi.org/10.1016/S0006-3207(02)00127-1

Clevenger, A. P., A. T. Ford, and M. A. Sawaya. 2009. Banff wildlife crossings project: integrating science and education in restoring population connectivity across transportation corridors. Final report to Parks Canada Agency, Radium Hot Springs, Canada. https://westerntransportationinstitute.org/research_projects/banff-wildlife-crossings-project-maintaining-ecological-connectivity-across-transportation-corridors-in-mountain-landscapes-utc/

Clevenger, A. P., A. R. Hardy, and K. Gunson. 2006. Analyses of wildlife-vehicle collision data: applications for guiding decision-making for wildlife crossing mitigation and motorist safety. I. Limiting effects of road-kill reporting data due to spatial inaccuracy. Unpublished report for the National Cooperative Highway Research Program, Utah State University, Logan, Utah, USA. https://westerntransportationinstitute.org/wp-content/uploads/2016/08/4W0518_Analysis_Report.pdf

Coelho, A. V. P., I. P. Coelho, F. T. Teixeira, and A. Kindel. 2014. Siriema: road mortality software. User’s manual v.2.0. Núcleo de Ecologia de Rodovias e Ferrovias, Universidade Federal do Rio Grande do Sul, Porto Alegre, Brazil. http://dx.doi.org/10.13140/RG.2.2.19953.38242

Darlington, S., A. Ladle, A. C. Burton, J. P. Volpe, and J. T. Fisher. 2022. Cumulative effects of human footprint, natural features and predation risk best predict seasonal resource selection by white-tailed deer. Scientific Reports 12:1072. https://doi.org/10.1038/s41598-022-05018-z

Dawe, K. L., E. M. Bayne, and S. Boutin. 2014. Influence of climate and human land use on the distribution of white-tailed deer (Odocoileus virginianus) in the western boreal forest. Canadian Journal of Zoology 92(4):353-363. https://doi.org/10.1139/cjz-2013-0262

Donker, S., and G. Chapman. 2015. Aerial ungulate survey (2015) for moose and white-tailed deer in WMU 503 (Lac La Biche). Alberta Environment and Parks-Lower Athabasca Region and Alberta Environmental Monitoring, Evaluation and Reporting Agency, Edmonton, Canada. http://osipfiles.alberta.ca/datasets/680/Aerial%20Ungulate%20Survey,%202015%20-%20WMU%20503.pdf

Dulac, J. 2013. Global land transport infrastructure requirements: estimating road and railway infrastructure capacity and costs to 2050. International Energy Agency, Paris, France. https://www.ctc-n.org/resources/global-land-transport-infrastructure-requirements-estimating-road-and-railway

Fahrig, L., and T. Rytwinski. 2009. Effects of roads on animal abundance: an empirical review and synthesis. Ecology and Society 14(1):21. https://doi.org/10.5751/ES-02815-140121

Finder, R. A., J. L. Roseberry, and A. Woolf. 1999. Site and landscape conditions at white-tailed deer/vehicle collision locations in Illinois. Landscape and Urban Planning 44(2-3):77-85. https://doi.org/10.1016/S0169-2046(99)00006-7

Fisher, J. T., A. C. Burton, L. Nolan, and L. Roy. 2020. Influences of landscape change and winter severity on invasive ungulate persistence in the Nearctic boreal forest. Scientific Reports 10:8742. https://doi.org/10.1038/s41598-020-65385-3

Ford, A. T., B. Dorsey, T. S. Lee, and A. P. Clevenger. 2022. A before-after-control-impact study of wildlife fencing along a highway in the Canadian Rocky Mountains. Frontiers in Conservation Science 3:935420. https://doi.org/10.3389/fcosc.2022.935420

Fuller, H. W., S. Frey, and J. T. Fisher. 2023. Integration of aerial surveys and resource selection analysis indicates human land use supports boreal deer expansion. Ecological Applications 33(1):e2722. https://doi.org/10.1002/eap.2722

Government of Alberta. 2020a. Aerial wildlife survey report: wildlife management unit 212 South East (2020). Government of Alberta, Edmonton, Canada. https://open.alberta.ca/publications/wildlife-management-unit-212-south-east-2020

Government of Alberta. 2020b. Annual report: transportation: 2019–2020. Government of Alberta, Edmonton, Alberta, Canada. https://open.alberta.ca/dataset/d24ecf3f-d15a-4e41-b090-9e1a4c85949a/resource/d170eb94-9440-46c3-ace8-c0939f3ba30a/download/trans-annual-report-2019-2020.pdf

Groot Bruinderink, G. W. T. A., and E. Hazebroek. 1996. Ungulate traffic collisions in Europe. Conservation Biology 10(4):1059-1067. https://doi.org/10.1046/j.1523-1739.1996.10041059.x

Gunson, K. E., A. P. Clevenger, A. T. Ford, J. A. Bissonette, and A. Hardy. 2009. A comparison of data sets varying in spatial accuracy used to predict the occurrence of wildlife-vehicle collisions. Environmental Management 44:268-277. https://doi.org/10.1007/s00267-009-9303-y

Gunson, K. E., G. Mountrakis, and L. J. Quackenbush. 2011. Spatial wildlife-vehicle collision models: a review of current work and its application to transportation mitigation projects. Journal of Environmental Management 92(4):1074-1082. https://doi.org/10.1016/j.jenvman.2010.11.027

Hubbard, M. W., B. J. Danielson, and R. A. Schmitz. 2000. Factors influencing the location of deer-vehicle accidents in Iowa. Journal of Wildlife Management 64(3):707-713. https://doi.org/10.2307/3802740

Huber, M. N. 2022. How the qualities of actor-issue interdependencies influence collaboration patterns. Ecology and Society 27(4):6 https://doi.org/10.5751/ES-13536-270406

Huijser, M. P., and J. S. Begley. 2022. Implementing wildlife fences along highways at the appropriate spatial scale: a case study of reducing road mortality of Florida Key deer. Nature Conservation 47:283-302. https://doi.org/10.3897/natureconservation.47.72321

Huijser, M. P., and P. J. M. Bergers. 2000. The effect of roads and traffic on hedgehog (Erinaceus europaeus) populations. Biological Conservation 95(1):111-116. https://doi.org/10.1016/S0006-3207(00)00006-9

Huijser, M. P., J. W. Duffield, A. P. Clevenger, R. J. Ament, and P. T. McGowen. 2009. Cost-benefit analyses of mitigation measures aimed at reducing collisions with large ungulates in the United States and Canada: a decision support tool. Ecology and Society 14(2):15. http://www.ecologyandsociety.org/vol14/iss2/art15/

Huijser, M. P., E. R. Fairbank, W. Camel-Means, J. Graham, V. Watson, P. Basting, and D. Becker. 2016. Effectiveness of short sections of wildlife fencing and crossing structures along highways in reducing wildlife-vehicle collisions and providing safe crossing opportunities for large mammals. Biological Conservation 197:61-68. https://doi.org/10.1016/j.biocon.2016.02.002

Huijser, M. P., P. McGowen, J. Fuller, A. Hardy, A. Kociolek, A. P. Clevenger, D. Smith, and R. Ament. 2008. Wildlife-vehicle collision reduction study: report to Congress. FHWA-HRT-08-034. U.S. Department of Transportation, Federal Highway Administration, Washington, D.C., USA. https://www.fhwa.dot.gov/publications/research/safety/08034/

Hussain, A., J. B. Armstrong, D. B. Brown, and J. Hogland. 2007. Land-use pattern, urbanization, and deer-vehicle collisions in Alabama. Human-Wildlife Conflicts 1(1):89-96. https://www.jstor.org/stable/24875057

Ibisch, P. L., M. T. Hoffman, S. Kreft, G. Pe’er, V. Kati, L. Biber-Freudenberger, D. A. DellaSala, M. M. Vale, P. R. Hobson, and N. Selva. 2016. A global map of roadless areas and their conservation status. Science 354(6318):1423-1427. https://doi.org/10.1126/science.aaf7166

Ignatavičius, G., A. Ulevičius, V. Valskys, G. Trakimas, L. Galinskaitė, and P. E. Busher. 2020. Termporal patterns of ungulate-vehicle collisions in a sparsely populated country. European Journal of Wildlife Research 66:58. https://doi.org/10.1007/s10344-020-01396-9

Jacobson, S. L., L. L. Bliss-Ketchum, C. E. de Rivera, and W. P. Smith. 2016. A behavior-based framework for assessing barrier effects to wildlife from vehicle traffic volume. Ecosphere 7(4):e01345. https://doi.org/10.1002/ecs2.1345

Jaeger, J. A. G., J. Bowman, J. Brennan, L. Fahrig, D. Bert, J. Bouchard, N. Charbonneau, K. Frank, B. Gruber, and K. Tluk von Toschanowitz. 2005. Predicting when animal populations are at risk from roads: an interactive model of road avoidance behavior. Ecological Modelling 185(2-4):329-348. https://doi.org/10.1016/j.ecolmodel.2004.12.015

Keken, Z., J. Sedoník, T. Kušta, R. Andrášik, and M. Bíl. 2019. Roadside vegetation influences clustering of ungulate vehicle collisions. Transportation Research Part D: Transport and Environment 73:381-390. https://doi.org/10.1016/j.trd.2019.07.013

Latham, A. D. M., M. C. Latham, N. A. Mccutchen, and S. Boutin. 2011. Invading white-tailed deer change wolf–caribou dynamics in northeastern Alberta. Journal of Wildlife Management 75(1):204-212. https://doi.org/10.1002/jwmg.28

Lavsund, S., and F. Sandegren. 1991. Moose-vehicle relations in Sweden: a review. Alces: A Journal Devoted to the Biology and Management of Moose 27:118-126. https://www.alcesjournal.org/index.php/alces/article/view/1109

Lee, T. S., K. Rondeau, R. Schaufele, A. P. Clevenger, and D. Duke. 2021. Developing a correction factor to apply to animal-vehicle collision data for improved road mitigation measures. Wildlife Research 48(6):501-510. https://doi.org/10.1071/WR20090

Levine, N. 2010. CrimeStat: spatial statistics program for the analysis of crime incident locations (v 3.3). Ned Levine and Associates, Houston, Texas, and the National Institute of Justice, Washington, D.C., USA. https://nij.ojp.gov/topics/articles/crimestat-spatial-statistics-program-analysis-crime-incident-locations#copyright

Magle, S. B., L. S. Simoni, E. W. Lehrer, and J. S. Brown. 2014. Urban predator-prey association: coyote and deer distributions in the Chicago metropolitan area. Urban Ecosystems 17:875-891. https://doi.org/10.1007/s11252-014-0389-5

Meijer, J. R., M. A. J. Huijbregts, K. C. G. J. Schotten, and A. M. Schipper. 2018. Global patterns of current and future road infrastructure. Environmental Research Letters 13(6):064006. https://doi.org/10.1088/1748-9326/aabd42

Meister, S. R., M. M. Hing, W. G. M. Vanlaar, and R. D. Robertson. 2016. Road safety monitor 2014: driver behaviour and wildlife on the road in Canada. Traffic Injury Research Foundation, Ottawa, Canada. https://tirf.ca/projects/road-safety-monitor-driver-behaviour-wildlife-road-canada/

National Academies of Sciences, Engineering, and Medicine. 2007. Animal-vehicle collision data collection. National Academies Press, Washington, D.C., USA. https://doi.org/10.17226/23138

Ng, J. W., C. Nielsen, and C. C. St. Clair. 2008. Landscape and traffic factors influencing deer–vehicle collisions in an urban environment. Human-Wildlife Conflicts 2(1):34-47. https://www.jstor.org/stable/24875104

Paemelaere, E. A. D., A. Mejía, S. Quintero, M. Hallett, F. Li, A. Wilson, H. Barnabas, A. Albert, R. Li, L. Baird, G. Pereira, and J. Melville. 2023. The road towards wildlife friendlier infrastructure: mitigation planning through landscape-level priority settings and species connectivity frameworks. Environmental Impact Assessment Review 99:107010. https://doi.org/10.1016/j.eiar.2022.107010

Pagany, R. 2020. Wildlife-vehicle collisions - influencing factors, data collection and research methods. Biological Conservation, 251:108758. https://doi.org/10.1016/j.biocon.2020.108758

Plante, J., J. A. G. Jaeger, and A. Desrochers. 2019. How do landscape context and fences influence roadkill locations of small and medium-sized mammals? Journal of Environmental Management 235:511-520. https://doi.org/10.1016/j.jenvman.2018.10.093

Roedenbeck, I. A., L. Fahrig, C. S. Findlay, J. E. Houlahan, J. A. G. Jaeger, N. Klar, S. Kramer-Schadt, and E. A. van der Grift. 2007. The Rauischholzhausen agenda for road ecology. Ecology and Society 12(1):11. http://dx.doi.org/10.5751/ES-02011-120111

Rolandsen, C. M., E. J. Solberg, I. Herfindal, B. Van Moorter, and B.-E. Saether. 2011. Large-scale spatiotemporal variation in road mortality of moose: Is it all about population density? Ecosphere 2(10):1-12. https://doi.org/10.1890/ES11-00169.1

Santos, R. A. L., S. M. Santos, M. Santos-Reis, A. Picanço de Figueiredo, A. Bager, L. M. S. Aguiar, and F. Ascensão. 2016. Carcass persistence and detectability: reducing the uncertainty surrounding wildlife-vehicle collision surveys. PLoS ONE 11(11):e0165608. https://doi.org/10.1371/journal.pone.0165608

Schwartz, A. L. W., F. M. Shilling, and S. E. Perkins. 2020. The value of monitoring wildlife roadkill. European Journal of Wildlife Research 66:18. https://doi.org/10.1007/s10344-019-1357-4

Seiler, A. 2004. Trends and spatial patterns in ungulate-vehicle collisions in Sweden. Wildlife Biology 10(4):301-313. https://doi.org/10.2981/wlb.2004.036

Seiler, A. 2005. Predicting locations of moose–vehicle collisions in Sweden. Journal of Applied Ecology 42(2):371-382. https://doi.org/10.1111/j.1365-2664.2005.01013.x

Seiler, A., and J. O. Helldin. 2006. Mortality in wildlife due to transportation. Pages 165-189 in J. Davenport and J. L. Davenport, editors. The ecology of transportation: managing mobility for the environment. Springer, Dordrecht, The Netherlands. https://doi.org/10.1007/1-4020-4504-2_8

Shilling, F., W. Collinson, M. Bil, D. Vercayie, F. Heigl, S. E. Perkins, and S. MacDougall. 2020. Designing wildlife-vehicle conflict observation systems to inform ecology and transportation studies. Biological Conservation 251:108797. https://doi.org/10.1016/j.biocon.2020.108797

Shilling, F., T. Nguyen, M. Saleh, M. K. Kyaw, K. Tapia, G. Trujillo, M. Bejarano, D. Waetjen, J. Peterson, G. Kalisz, R. Sejour, S. Croston, and E. Ham. 2021. A reprieve from US wildlife mortality on roads during the COVID-19 pandemic. Biological Conservation 256:109013. https://doi.org/10.1016/j.biocon.2021.109013

Shilling, F. M., and D. P. Waetjen. 2015. Wildlife-vehicle collision hotspots at US highway extents: scale and data source effects. Nature Conservation 11:41-60. https://doi.org/10.3897/natureconservation.11.4438

Snow, N. P., W. F. Porter, and D. M. Williams. 2015. Underreporting of wildlife-vehicle collisions does not hinder predictive models for large ungulates. Biological Conservation 181:44-53. https://doi.org/10.1016/j.biocon.2014.10.030

Steenberghen, T., K. Aerts, and I. Thomas. 2010. Spatial clustering of events on a network. Journal of Transport Geography 18(3):411-418. https://doi.org/10.1016/j.jtrangeo.2009.08.005

Steiner, W., F. Leisch, and K. Hackländer. 2014. A review on the temporal pattern of deer–vehicle accidents: impact of seasonal, diurnal and lunar effects in cervids. Accident Analysis and Prevention 66:168-181. https://doi.org/10.1016/j.aap.2014.01.020

Teixeira, F. Z., T. Rytwinski, and L. Fahrig. 2020. Inference in road ecology research: what we know versus what we think we know. Biology Letters 16(7):20200140. https://doi.org/10.1098/rsbl.2020.0140

van der Ree, R., J. A. G. Jaeger, E. A. van der Grift, and A. P. Clevenger. 2011. Effects of roads and traffic on wildlife populations and landscape function: road ecology is moving towards larger scales. Ecology and Society 16(1):48. http://dx.doi.org/10.5751/ES-03982-160148

Vanlaar, W. G. M., H. Barrett, M. Mainegra Hing, S. W. Brown, and R. D. Robertson. 2019. Canadian wildlife-vehicle collisions: an examination of knowledge and behavior for collision prevention. Journal of Safety Research 68:181-186. https://doi.org/10.1016/j.jsr.2018.12.003

Visintin, C., R. van der Ree, and M. A. McCarthy. 2016. A simple framework for a complex problem? Predicting wildlife–vehicle collisions. Ecology and Evolution 6(17):6409-6421. https://doi.org/10.1002/ece3.2306

Weiss, D. J., A. Nelson, H. S. Gibson, W. Temperley, S. Peedell, A. Lieber, M. Hancher, E. Poyart, S. Belchior, N. Fullman, B. Mappin, U. Dalrymple, J. Rozier, T. C. D. Lucas, R. E. Howes, L. S. Tusting, S. Y. Kang, E. Cameron, D. Bisanzio, K. E. Battle, S. Bhatt, and P. W. Gething. 2018. A global map of travel time to cities to assess inequalities in accessibility in 2015. Nature 553:333-336. https://doi.org/10.1038/nature25181

Wilkins, D. C., K. M. Kockelman, and N. Jiang. 2019. Animal-vehicle collisions in Texas: how to protect travelers and animals on roadways. Accident Analysis and Prevention 131:157-170. https://doi.org/10.1016/j.aap.2019.05.030