Temporal Patterns and Hotspot Identification of Vehicle Collisions with the Roe Deer (Capreolus pygargus) on Jeju Island, South Korea

Seung-Yun Baek1 and Seong-Min Lee2*

1Graduate School of Agricultural Science, Tokyo University of Agriculture and Technology, Fuchu, Tokyo, Japan

2Research Institute for Agriculture and Life Sciences, Seoul National University, Seoul, Korea

ABSTRACT

Wildlife–vehicle collisions (WVCs) are increasing worldwide, posing a threat to both humans and wildlife in terms of human injury and loss of wildlife. Understanding the temporal and spatial non-randomness of WVCs is an essential process for effective WVC management. Roe deer (Capreolus pygargus) is the most prevalent wildlife species in WVCs on Jeju Island, South Korea. This study analyzed individual temporal patterns and spatial hotspots (clusters) based on roe deer–vehicle collisions (RDVC) on Jeju Island from 2014 to 2017. RDVCs occur frequently from July to October, and RDVCs are higher in male deer than with females in all seasons, except in winter. The clusters in the high-risk periods were more regionally distributed compared to the clusters in the low-risk periods. Regardless of the period, some road sections have been identified as crucial clusters that should be given priority for road fencing to mitigate human-deer conflicts. Our findings could also provide road administrators with practical options for management such as temporarily imposing vehicle speed limits and installing signboards.

Article Information

Received 30 November 2020

Revised 11 December 2020

Accepted 16 December 2020

Available online 05 April 2021

(early access)

Published 16 December 2021

Authors’ Contribution

S-YB and S-ML conceived this research. S-YB conducted the cluster analysis. S-ML conducted the statistical analysis. S-YB and S-ML wrote the first draft and contributed by revising the manuscript.

Key words

Deer-vehicle collisions, Hotspots, Capreolus pygargus, KDE+, Road ecology

DOI: https://dx.doi.org/10.17582/journal.pjz/20201130031155

* Corresponding author: ecologist@naver.com

0030-9923/2022/0001-0347 $ 9.00/0

Copyright 2022 Zoological Society of Pakistan

INTRODUCTION

Wildlife–vehicle collisions (WVCs) are one of the most representative issues of wildlife–human conflict worldwide. WVCs lead to high socio-economic costs as they lead to vehicle damage, human injury, or death, as well as a decrease in wildlife (Putman et al., 2004; Huijser et al., 2008; Gren and Jägerbrand, 2019). In the United States, it is estimated that 26,000 human injuries, 200 human deaths, and a cost of more than 8 billion USD result from the 1–2 million WVCs per year (Huijser et al., 2008). In Germany, approximately 200,000 the European roe deer (Capreolus capreolus Linnaeus, 1758) vehicle collisions result in 3,000 human injuries and 50 deaths, with a cost of 5 million euros (approximately 5.9 million USD) per year (Hothorn et al., 2012). Various attempts have been made to reduce WVCs (Huijser et al., 2009; Langbein et al., 2011). Fencing is recognized as the most effective mitigation measure (Glista et al., 2009). Fencing across the entire road acts as a physical barrier that limits wildlife movement (Jaeger and Fahrig, 2004), but it is also largely impossible due to financial constraints. Since WVCs occur non-randomly (Rodríguez-Morales et al., 2013), understanding the temporal and spatial clustering of WVCs is a fundamental step in effective mitigation measures. Since WVCs are closely related to the behavior and activities of each species (Clevenger et al., 2003; Pagany, 2020), there is generally an increased risk during the dispersal period and the rutting season (Haigh, 2012).

Deer–vehicle collisions (DVCs) are the most frequent type of WVC across Europe and the U.S. (Huijser et al., 2008; Sjölund, 2016; Bartonička et al., 2018). DVC from 2018 to 2019 and accounted for 52% of WVCs in South Korea from 2018 to 2019 (National Institute of Ecology, 2019). On Jeju Island, DVCs accounted for more than 90% of the WVCs (Jeju Special Self-Governing Province, 2018). As Jeju Island is one of South Korea’s most popular tourist destinations, with more than 10 million tourists visiting each year, safety threats to tourists by DVCs have been raised to a serious level. The Siberian roe deer (Capreolus pygargus Pallas, 1771), the only medium-sized, herbivorous, mammalian species that currently inhabits Jeju Island, was an endangered species in the 1980s. However, its population size has increased due to consistent conservation actions, causing more than 200 crop damage cases per year on Jeju Island. Roe deer on Jeju Island has been reported to be morphologically, genetically, and ecologically different from roe deer in other regions (Oh, 2004; Park et al., 2011; Lee et al., 2015). Therefore, it is necessary to collect field data on Jeju Island to ensure the safety of tourists and reduce roe deer–vehicle collisions (RDVCs). Thus, the purpose of this study was to identify individual temporal patterns and spatial hotspots (clusters) based on the period of RDVC occurrence risk by analyzing the RDVCs on Jeju Island, as well as to suggest efficient mitigation measures for RDVCs.

MATERIALS AND METHODS

Study area

Jeju Island (33°10’ N, 126°10’ E) is a volcanic island 85 km south of the Korean Peninsula. It is oval-shaped, with a total area of 1,850 km2, has Mt. Halla in the center, and is 73 km east to west and 41 km north to south. Forests (35%), croplands (35%), and grasslands (19%) are the most common land cover types on the island (Fig. 1). The average annual temperature is 16.9°C, while the lowest is -1.6°C and the highest is 37.0°C. The climate varies from cold in winter to humid monsoon in summer. The average annual rainfall is 2,061 mm, with the highest (at least 200 mm) monthly precipitation in August and the lowest in December (Chang et al., 2019). Depending on the altitude, the vegetation there is classified as evergreen broad-leaved forest (e.g., Quercus salicina, Q. glauca, and Castanopsis sieboldii) from the coast to 600 m above sea level (a.s.l.), deciduous broadleaved forest (e.g., Q. acutissima, Q. serrata, and Carpinus tschonoskii) between 600 to 1,400 m a.s.l., and subalpine coniferous forest (e.g., Abies koreana and Juniperus chinensis) that accompanies scrubs (e.g., Rhododendron mucronulatum and Empetrum nigrum var. japonicum) between 1,400 and 1,950 m a.s.l. (Chung, 2007).

 

Roe deer vehicle collisions data

We used RDVC data recorded from January 2014 to December 2017 by the Jeju Wildlife Rescue Center and from January 2015 to December 2017 by the local government (i.e., Jeju City and Seogwipo City). The data from the Jeju Wildlife Rescue Center included collision dates, GPS coordinates, and individual information (i.e., sex of animal and age). The data of the local government included only information on the collision date and location, excluding the individual information, due to the absence of professional personnel. In the recorded data, the cases in which location information was incorrectly described were excluded from the spatial analysis.

Individual-temporal analysis

In the temporal study, a total of 1,324 RDVCs were divided by month to analyze the monthly patterns. Using data from the Jeju Wildlife Rescue Center, which included individual information (i.e., sex and age), a temporal analysis of the individuals was performed. For this, periods were defined by the four seasons: spring (March–May), summer (June–August), autumn (September–November), and winter (December–February). Individuals were assigned an age category of fawn or adult. Since the behavioral differences between the sexes of deer occur after the natal dispersal period (when the fawns become independent of the mother), fawns were excluded from the sex category (male = 70, female = 44). A two-way Chi–square test was performed to analyze the individual temporal patterns.

Identification of clusters

The KDE+ method (Bíl et al., 2013) was applied to identify clusters of RDVCs. The KDE+ method is a kernel density estimation method that estimates the probability density function of the underlying data using kernel functions and has been used in many studies to identify statistically significant, road cluster sections (Sjölund, 2016; Bartonička et al., 2018; Favilli et al., 2018). As the KDE+ method is affected by the length of each road, number of points on each road, and distance between points, only RDVC points (n = 720) that occurred on major roads were used for analysis, excluding points on discontinuous and short roads (Fig. 4). The clusters were identified by distinguishing between the high- and low-risk periods of RDVC occurrence.

RESULTS

Individual-temporal patterns

RDVCs occurred at an average of 118 times per month, and from July to October, they occurred higher than the average (Fig. 2). In addition, the frequency of RDVCs between the females and males in each season was different (χ2 = 8.57, df = 3, P = 0.036; Fig. 3). The frequency of RDVCs with males was the highest in autumn, with 20 cases (31%), and the lowest in winter, with 12 cases (19%). On the other hand, the frequency of RDVCs with females was the highest in winter, with 17 cases (39%), and the lowest in autumn, with 8 cases (18%). There was no difference (χ2 = 5.97, df = 3, P = 0.113) between the seasonal fawn and female RDVC frequencies, but there was a difference (χ2 = 12.73, df = 3, P = 0.005) between the seasonal fawn and male RDVC frequencies (Fig. 3).

 

 

Clustering of risk period

Eighty-two clusters (median: 157.8 m, range: 66 to 438 m) and 49 clusters (median: 155.1 m, range: 41 to 380 m) were identified for the high-risk period (HRP; summer–autumn) and low-risk period (LRP; winter–spring) of RDVC occurrence (Fig. 4). Although there was no difference between the size of each cluster in both the periods (Mann–Whitney test: U = 1936, P = 0.7302), the HRP had more clusters than the LRP and occurred over a wider region (Fig. 4). Some road sections had clusters in both the periods, which were mostly located around the northeast, north, and west side of Mt. Halla.

 

DISCUSSION

In this study, RDVCs on Jeju Island varied based on the individuals involved and the period in which they occurred. The results also showed that spatial clusters of RDVC were generated according to risk period.

RDVC temporal pattern for individuals

RDVCs occurred most frequently between July and October, which included summer and autumn, and there were different temporal patterns, depending on sex and age. Male roe deer tend to establish and maintain their territory from spring to the rutting season, showing higher activity levels than females in the rutting season (Johansson, 1996; Oh, 2004; Pagon et al., 2017), as well as in other seasons, except for winter (Kämmerle et al., 2017). The relatively high RDVC occurrence in males appears to indicate a difference in these behaviors. In particular, RDVC occurrence was 3.25 higher in males than in females in the autumn, from September to October, when roe deer on Jeju Island were in the rutting season and males were mating while protecting their territory (Oh, 2004). These results are consistent with those shown in the vehicle collisions of European roe deer (Capreolus capreolus; Vincent et al., 1988), white-tailed deer (Odocoileus virginianus; Allen and McCullough, 1976), and moose (Alces alces; Belant, 1995). July, the month with the highest frequency of RDVCs, is the period when fawns begin to move around with their mothers (Oh, 2004). Groot-Bruinderink and Hazbroek (1996) revealed that European roe deer had a high frequency of vehicle collisions in late spring and summer in relation to breeding and dispersion. Langbein and Putman (2006) described that the period when European roe deer mothers move with their fawns could expose them more to the threat of vehicle collisions, and according to Tufto et al. (1996), a lactating female travels longer distances than a non-lactating female to improve the nutrition of their milk. In addition, data from the Jeju Wildlife Rescue Center have shown that the number of RDVCs with fawns in summer is relatively high, accounting for 38% of the summer RDVCs, and that there is no significant difference between the frequency of seasonal collisions between adult females and fawns; therefore, the RDVCs during this period are likely to be affected by the activities of lactating females and fawns. Thus, our results indicate that RDVCs on Jeju may be affected by the temporal behavior of roe deer.

RDVC clusters

We identified RDVC clusters using the KDE+ method and demonstrated that clusters varied by risk period. WVCs are affected by species-specific factor (Clevenger et al., 2003; Gunson et al., 2011; Pagany, 2020) and deer behavior changes are driven by food availability, climate, and phenology (Morellet et al., 2009; Bischof et al., 2012).

We found that the RDVCs were affected by the periodic activity patterns of roe deer and were spatially clustered. The findings from our study can be used by policymakers or road administrators as a basis for cost-effective mitigation measures. As there is a temporal difference between the risk and clusters, it is expected that if preferential measures are taken for clusters during a period when the RDVCs occur at a higher frequency, the RDVCs can be effectively controlled. As some road sections have been identified as cluster sections regardless of the period, we suggest that these sections should be prioritized for road fencing. In addition, as WVCs are known to be affected by vehicle speed and volume, as well as by wildlife ecological factors (Seiler, 2005; Lao et al., 2011; Arevalo et al., 2017), the installation of temporary signs to alert drivers to risks (especially from July to October) and imposing vehicle speed limits are expected to mitigate RDVCs. Davenport and Davenport (2006) mention that WVCs are affected by complex interactions between animal ecology, transportation, and environmental factors. In the future study, more detailed and systematic approaches are needed that includes traffic and environmental factors that could not be analyzed in this study.

We conclude that spatio-temporal RDVCs and clustering related to ecological factors impact RDVCs roe deer conflict. The cluster road sections should be considered as a priority management strategy to effectively reduce RDVCs.

ACKNOWLEDGEMENTS

We are grateful to the Jeju Wildlife Rescue Center and Jeju-si City hall and Seogwipo-si City hall for their support data for this research project. We are also deeply grateful to Jirí Sedoník for his help in running the KDE+ cluster analysis.

Statement of conflict of interest

The authors have declared no conflict of interest.

REFERENCES

Allen, R.E. and McCullough, D.R., 1976. Deer–car accidents in southern Michigan. J. Wildl. Manage., 40: 317-325. https://doi.org/10.2307/3800431

Arevalo, J.E., Honda, W., Arce-Arias, A. and Häger, A., 2017. Spatiotemporal variation of road kills shows mass mortality events for amphibians in a highly trafficked road adjacent to a national park, Costa Rica. Rev. Biol. Trop., 65: 1261. https://doi.org/10.15517/rbt.v65i4.27903

Bartonička, T., Andrášik, R., Duľa, M., Sedoník, J. and Bíl, M., 2018. Identification of local factors causing clustering of animal-vehicle collisions. J. Wildl. Manage., 82: 940–947. https://doi.org/10.1002/jwmg.21467

Belant, J.L., 1995. Moose collisions with vehicles and trains in northeastern Minnesota. Alces, 31: 45–52.

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

Bischof, R., Loe, L.E., Meisingset, E.L., Zimmermann, B., Van-Moorter, B. and Mysterud, A., 2012. A migratory northern ungulate in the pursuit of spring: jumping or surfing the green wave? Am. Nat., 180: 407–424. https://doi.org/10.1086/667590

Chang, S.W., Chung, I., Kim, M., Tolera, M. and Koh, G., 2019. Application of GALDIT in assessing the seawater intrusion vulnerability of Jeju Island, South Korea. Water, 11: 1824. https://doi.org/10.3390/w11091824

Chung, C.H., 2007. Vegetation response to climate change on Jeju Island, South Korea, during the last deglaciation based on pollen record. Geosci. J., 11: 147–155. https://doi.org/10.1007/BF02913928

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

Davenport, J. and Davenport, J.L., 2006. The ecology of transportation: Managing mobility for the environment. Springer, Netherlands. https://doi.org/10.1007/1-4020-4504-2

Favilli, F., Bíl, M., Sedoník, J., Andrášik, R., Kasal, P., Agreiter, A. and Streifeneder, T., 2018. Application of KDE+ software to identify collective risk hotspots of ungulate-vehicle collisions in South Tyrol, northern Italy. Eur. J. Wildl. Res., 64: 59. https://doi.org/10.1007/s10344-018-1214-x

Glista, D.J., DeVault, T.L. and DeWoody, J.A., 2009. A review of mitigation measures for reducing wildlife mortality on roadways. Landsc. Urban Pl., 91: 1–7. https://doi.org/10.1016/j.landurbplan.2008.11.001

Gren, I.M. and Jägerbrand, A.K., 2019. Calculating the costs of animal-vehicle accidents involving ungulate in Sweden. Transp. Res. Part D, Transp. Environ., 70: 112-122. https://doi.org/10.1016/j.trd.2019.03.008

Groot-Bruinderink, G.W.T.A. and Hazebroek, E., 1996. Ungulate traffic collisions in Europe. Conserv. Biol., 10: 1059–1067. https://doi.org/10.1046/j.1523-1739.1996.10041059.x

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

Haigh, A.J., 2012. Annual patterns of mammalian mortality on Irish roads. Hystryx, 23: 58–66.

Hothorn, T., Brandl, R. and Müller, J., 2012. Large-scale model-based assessment of deer-vehicle collision risk. PLoS ONE, 7: e29510. https://doi.org/10.1371/journal.pone.0029510

Huijser, A.P., McGowen, P., Clevenger, A.P. and Ament, R., 2008. Wildlife collision reduction study: Best practices manual. Washington, D.C., USA.

Huijser, M.P., Duffield, J.W., Clevenger, A.P., Ament, R.J. and McGowen, P.T., 2009. Cost-benefit analyses of mitigation measures aimed at reducing collisions with large ungulates in the United States and Canada: A decision support tool. Ecol. Soc., 14: 15. https://doi.org/10.5751/ES-03000-140215

Jaeger, J.A.G. and Fahrig, L., 2004. Under what conditions do fences reduce the effects of roads on population persistence? Conserv. Biol., 18:1651–1657. https://doi.org/10.1111/j.1523-1739.2004.00304.x

Jeju Special Self-Governing Province, 2018. Jeju Road-kill Report. Jeju, Korea.

Johansson, A., 1996. Territory establishment and antler cycle in male roe deer. Ethology, 102: 549–559. https://doi.org/10.1111/j.1439-0310.1996.tb01147.x

Kämmerle, J.-L., Brieger, F., Kröschel, M., Hagen, R., Storch, I. and Suchant, R., 2017. Temporal patterns in road crossing behaviour in roe deer (Capreolus capreolus) at sites with wildlife warning reflectors. PLoS One, 12: e0184761. https://doi.org/10.1371/journal.pone.0184761

Langbein, J. and Putman, R., 2006. National Deer-Vehicle Collisions Project: Scotland (2003–2005). Wrexham, UK.

Langbein, J., Putman, R., Pokorny, B., 2011. Traffic collisions involving deer and other ungulates in Europe and available measures for mitigation. In: Ungulate management in Europe: Problems and practices (eds. R. Putman, M. Apollonio and R. Andersen). Cambridge Univ, UK. pp. 215–259. https://doi.org/10.1017/CBO9780511974137.009

Lao, Y., Wu, Y.-J., Corey, J. and Wang, Y., 2011. Modeling animal-vehicle collisions using diagonal inflated bivariate Poisson regression. Accid. Anal. Prev., 43: 220–227. https://doi.org/10.1016/j.aap.2010.08.013

Lee, Y.S., Markov, N., Voloshina, I., Argunov, A., Bayarlkhagva, D., Oh, J.G., Park, Y.S., Min, M.S., Lee, H. and Kim, K.S., 2015. Genetic diversity and genetic structure of the Siberian roe deer (Capreolus pygargus) populations from Asia. BMC Genet., 16: 100. https://doi.org/10.1186/s12863-015-0244-6

Morellet, N., Verheyden, H., Angibault, J.M., Cargnelutti, B., Lourtet, B. and Hewison, A.J.M., 2009. The effect of capture on ranging behaviour and activity of the European roe deer Capreolus capreolus. Wildl. Biol., 15: 278–287. https://doi.org/10.2981/08-084

National Institute of Ecology, 2019. Intensive Survey on Road-kill Hotspots in South Korea. National Institute of Ecology, Korea.

Oh, J.G., 2004. Characteristics of ecological behaviour of roe deer (Capreolus pygargus tianschanicus) in jeju island Korea. PhD thesis. Korea National University of Education, Cheongju-si, Korea.

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

Pagon, N., Grignolio, S., Brivio, F., Marcon, A. and Apollonio, M., 2017. Territorial behaviour of male roe deer: a telemetry study of spatial behaviour and activity levels. Folia Zool., 66: 267–276. https://doi.org/10.25225/fozo.v66.i4.a9.2017

Park, Y.S., Lee, W.S., Kim, J.T. and Oh, H.S., 2011. Morphological examination of the siberian roe deer Capreolus pygargus in South Korea. J. Anim. Vet. Adv., 10: 2874-2878.

Putman, R.J., Langbein, J. and Staines, B.W., 2004. Deer and road traffic accidents: A review of mitigation measures: Costs and cost-effectiveness. Inverness, Scotland.

Rodríguez-Morales, B., Díaz-Varela, E.R. and Marey-Pérez, M.F., 2013. Spatiotemporal analysis of vehicle collisions involving wild boar and roe deer in nw spain. Accid. Anal. Prev., 60: 121–133. https://doi.org/10.1016/j.aap.2013.07.032

Seiler, A., 2005. Predicting locations of moose–vehicle collisions in Sweden. J. appl. Ecol., 42: 371–382. https://doi.org/10.1111/j.1365-2664.2005.01013.x

Sjölund, M., 2016. Road and landscape features affecting the aggregation of ungulate vehicle collisions in southern Sweden. Master dissertation, Swedish University of Agricultural Sciences, Uppsala, Sweden.

Tufto, J., Andersen, R. and Linnell, J.D.C., 1996. Habitat use and ecological correlates of size in a small cervid: the roe deer. J. Anim. Ecol., 65: 715–724. https://doi.org/10.2307/5670

Vincent, J.P., Bideau, E., Cibien, C. and Quéré, J.P., 1988. Traffic deaths in roe deer (Capreolus capreolus) - example of woodland area in the Paris Basin. Z. Jagdwissen., 34: 63-68. https://doi.org/10.1007/BF02241282