Introduction

Since the majority of wildlife habitats are fragmented by linear structures, the occurrence of unwanted dangerous encounters that happen on roads worldwide is common. Roadkill is among the most significant threats of wildlife and humans safety. Annually, two million vertebrate and 600 thousand mammal roadkills are estimated to occur in the Republic of Korea (Choi, 2016), indicating the urgent need for roadkill mitigation plans for humans and wildlife.

Numerous factors make roads more prone to roadkills. Parameters linked to road composition, such as, road width, curvature, and structure, affect the spatial patterns of roadkills (Byun et al., 2016; Clevenger & Kociolek, 2013; Kim et al., 2019a). In particular, road structures that interfere with the behavior of wildlife, such as, road crossing, could be a critical factor in determining the spatial patterns of roadkills. For example, wildlife fencing is considered as one of the most effective roadkill mitigation measures, as it prevents wildlife from invading roads (Rytwinski et al., 2016). However, when fence length is not long enough, roadkill risk significantly increases at fence ends (Plante et al ., 2019). Various types of fencing structures at the road edge, such as, rock fall fences or noise fences, have similar effects. On the other hand, obstacles with lower heights at road edges, such as, guard rails, do not significantly contribute to avoiding wildlife from invading roads (Byun et al., 2016).

When animals are already on the road other types of barriers can prevent them from crossing the road. For example, the median barriers reduce the permeability of roads to wildlife and the barrier effect increase roadkill risk (Clevenger & Kociolek, 2013). The median barrier may exhibit different effects depending on the body size of the animal and the structure or material of the barrier. in addition, visibility also influences roadkill (Collinson, 2013; Ignatavicius & Valskys, 2018; Kioko et al., 2015; Smith-Patten & Patten, 2008). Reduced visibility on roadways at night is considered dangerous to nocturnal wildlife (Braunstein, 1998). Therefore, understanding how these factors influence the spatial characteristics of roadkills is crucial when establishing mitigation plans (Kim et al ., 2019a).

A total of 21,397 roadkills were collected from the Korea Roadkill Observation System (KROS) in the Republic of Korea in 2019 (Kim et al., 2019b). Based on these data, Song et al. (2020) analyzed 50 roadkill hotpots on national highways throughout the country using the kernel density estimate with 1 km search radius. Subsequently, the 50 hotspots were hierarchically divided into five categories, from rank one (low density) to five (high density) according to the z-value. In the present study, we surveyed the 50 roadkill hot spots previously identified by Song et al. (2020) and collected data on nine types of road structures which might influence roadkill risk. Our study aimed to identify road structures which might affect the risk of roadkills in order to suggest future roadkill mitigation plans in the Republic of Korea.

Materials and Methods

We surveyed the previously identified 50 roadkill hotspots which were ranked from one to five according to roadkill density from April to June 2020 (Fig. 1). Of the 50 hotspots, 4, 8, 8, 11, and 19 hotspots were ranked as one, two, three, four, and five, respectively. The average number of cases involving roadkill was 2.7, 4.1, 5.9, 13.2, and 23.1 cases per kilometer for rank one to five respectively (www.nie-ecobank.kr). We collected data on the pre-

새창으로 보기
Table 1.
Nine types of road structures and descriptions recorded from roadkill hotspot field survey
Category Road structure Abbreviation Description
Median Median barrier MB Longitudinal barrier that distinguish two side of traffic in the middle of roads
Bottom blocked median barrier MB_BB
Edge_low Guard rail GR Longitudinal barrier that keeps vehicles within the roads
Bottom blocked guard rail GR_BB
Edge_high Rock fall fence RF Longitudinal barrier that protects road by preventing rocks fall into road way
Noise fence NF Longitudinal barrier to reduce noise produced from road way
Visibility Light LG Keeps roadway bright when dark
Bridge Bridge BR A structure built to span a physical obstacle without blocking the way underneath

sence and absence of eight types of road structures, namel y, median barrier, bottom blocked median barrier, guard rail, bottom blocked guard rail, rock fall fence, noise fence, light, and bridge. The data were classified into five categories according to their characteristics for statistical analyses (Table 1, Fig. 2).

새창으로 보기
Fig. 1.
Map of the study area. Five ranked road-kill hotspot are presented differently by each colour.
PNIE2-4-6-f1.jpg
새창으로 보기
Fig. 2.
Typical form of (a) bottom blocked median barrier, (b) bottom blocked guard rail, (c) rock fall fence (d) noise fence, (e) light and (f) bridge.
PNIE2-4-6-f2.jpg

To investigate the effects of road structures on the probability of roadkills, we conducted a cumulative link model (CLM) using the function CLM from the ordinal package. We used the roadkill rank (one–five) as the variable response. Among the 50 roadkill hotspots, median barrier and guard rail were present in 46 and 48 sections, respectively and were therefore removed from the analysis due to low representation. To test the effects of barriers at the center and edge of the road, the presence of bottom blocked median barrier and bottom blocked guard rail were included in the model as predictors named ‘median’ and ‘edge_low’ (Table 1). We considered the presence of either noise fence or rock fall fence as a high barrier and included the presence of a high barrier (presence of noise fence or rock fall fence) in the model as the predictor named ‘edge_ high’ (Table 1). Furthermore, we included light in the model as a predictor to test the effect of visibility-related structure named ‘visibility’ (Table 1). Finally, we included bridge in the model as a predictor named "bridge" to test how disconnection between road surface and surrounding landscape might influence roadkill rank (Table 1).

Results and Discussion

In general, 60 % of the roadkill hotspot rank one had bottom blocked median barrier, whereas the other ranks included a lower percentage (12.5 % in rank two, 14.3 % in rank three, 0 % in rank four, and 6.3% in rank five) (Table 2). When visually investigating the patterns between road structure sand roadkill ranks, bottom blocked guard rail was negatively correlated to roadkill hotspot rank (50 % in rank two, 28.6 % in rank three, and 14.3 % in rank four), but no bottom blocked guard rail was observed in ranks one and five (Table 2). The highest percentage of noise fence was observed in roadkill hotspot rank two (62.5 %) and the lowest percentage was found in rank three (14.3 %), without any positive or negative tendency. The percentage of rock fall fence was the highest in rank five (43.8 %) and the lowest in rank one (0 %). Light was the most frequently observed in rank three (42.9 %) and was the lowest in rank four (21.4 %). Finally, the percentage of hotspot rank with bridge was the highest in hotspot rank five (43.8 %), however, no particular tendency was observed (Table 2).

새창으로 보기
Table 2.
Percentage of each road structure type on each roadkill hotspot rank. The number of each roadkill hotspot rank is 5 for rank 1, 8 for rank 2, 7 for rank 3, 14 for rank 4, and 16 for rank 5
MB_BB (%) GR_BB (%) RF (%) NF (%) SC (%) TL (%) LG (%) BR (%)
Rank 1 60.0 0.0 0.0 40.0 0.0 20.0 40.0 0.0
Rank 2 12.5 50.0 37.5 62.5 37.5 0.0 25.0 37.5
Rank 3 14.3 28.6 14.3 14.3 28.6 28.6 42.9 42.9
Rank 4 0.0 14.3 42.9 35.7 42.9 21.4 21.4 35.7
Rank 5 6.3 0.0 43.8 43.8 25.0 25.0 31.3 43.8
Average 18.6 18.6 27.7 39.3 26.8 19.0 32.1 32.0

The model for investigating the effects of road structures on roadkill rank was significant compared to the null model (x2=12.416, df=5, p=0.029). The roadkill rank was significantly (p=0.026) higher in the absence of median category (Table 3, Fig. 3). No effect on the other parameters of the roadkill rank was observed (Table 3).

새창으로 보기
Table 3.
The effects of road structures on the roadkill rank using cumulative link model
Parameter Estimate S.E z value P
Median (MB_BB) -2.236 1.003 -2.229 0.026
Edge_low (GR_BB) -1.112 0.742 -1.500 0.134
Edge_high (RF+NF) -0.349 0.587 -0.594 0.552
Visibility (LG) -0.321 0.598 -0.536 0.592
Bridge (BR) 1.095 0.608 1.800 0.072
새창으로 보기
Fig. 3.
Estimated marginal means depending on the absence/presence of MB_BB, GR_BB, EDGE_high, Visibility, and bridge. It indicates when the estimated marginal mean is higher, the road kill probability is higher. The bars indicate confident interval in 95%.
PNIE2-4-6-f3.jpg

Although the model that contained all five road structure categories indicated significance, besides median, other parameters were not significantly affecting the roadkill rank. Similar to our result, other studies also have revealed that animals were less likely to cross roads when median barriers were present (Barnum, 2003; Gunson et al ., 2011; Malo et al., 2004). In contrast, Clevenger and Kociolek (2013) concluded that low permeability due to median structures on roads increased the risk of roadkill. In the Republic of Korea, the standard size of the concrete median barrier is 140.6 cm (Kim et al., 2009). According to a previous study conducted by Park et al. (2018), the deterrence rate at a wildlife fence of 140 cm was 83.3 % with respect to water deer (Hydropotes inermis), which is the most frequent victim of roadkills in the Republic of Korea. Thus, the study concluded that median barrier decreased road crossing attempts by wildlife and roadkill risk.

Even though our result did not reveal the significant effect of bottom blocked guard rail, we reported a negative tendency of bottom blocked guard rail with respect to roadkill rank (Table 3, Fig. 3). Song et al. (2011) also reported that a two-layer guard rail with a height of 90 cm, referred to as bottom blocked guard rail in this study, was effective in preventing roadkills on an expressway. In contrast, Byun et al. (2016) concluded that the guard rail had little contribution in reducing roadkills. In addition, it was reported that, in the Republic of Korea, the deterrence rate of a fence below 100 cm was zero with respect to water deer (Park et al., 2018). Thus, there is need for more detailed study concerning the effect of guard rail or bottom blocked guard rail on roadkill mitigation. Furthermore, our result also showed that roadkill probability decreased in the presence of bridges (Table 3, Fig. 3). So far, many studies have been focusing on wild life crossings which connect both sides of fragmented habitats. These types of bridges help wild animals to avoid roadkills and enabled safe crossing of roads (Williams et al., 2019). Bridge disconnects road surface and surrounding landscape, thereby excluding wild animals from roads and lowering roadkill risk.

Our research showed that the existence of bottom blocked median barrier decreased roadkill risk. Spatial patterns of roadkill depends on various factors, such as, wildlife density, seasonality, and landscape change (Kim et al., 2019a; Saeki & Macdonald, 2004; Seo et a l., 2015). Thus, we believe that consistent monitoring and analysis are necessary to respond to the changes in roadkill hotspots. For further studies, we recommend researchers to investigate how road structures affect roadkill risk by comparing roadkill hotspots and coldspots, which refers to road sections with low roadkill density. In addition, quantitative analysis between the number of roadkills and the presence of road structures may also give an insight for management of roadkill mitigation plans.

Conflict of Interest

The authors declare that they have no competing interests.

Acknowledgments

This work was supported by a grant from the National Institute of Ecology (NIE), funded by the Ministry of Environment (MOE) of the Republic of Korea (NIE-C-2021-28).

References

1 

Barnum S.A. ( 2003 ). Identifying the best locations along highways to provide safe crossing opportunities for wildlife . Colorado Department of Transportation. Research Branch, Report No. CDOT-DTD-UCD-2003-9 .

2 

Braunstein, M.M. (1998). Roadkill: Driving animals to their graves. Animal Issues, 29(3), 22-28. Retrieved July 30, 2021 from , http://www.culturechange.org/issue8/roadkill.htm .

3 

Byun Y.S , Kwon J.N , Kim J.H , Shin M.H , Lee S.D. ( 2016 ). How Do landscape and road barriers affect road crossing of multihabitat mammals . Journal of the Korean Association of Geographic Information Studies, 19(3), 89-101 , doi:10.11108/kagis.2016.19.3.089 .

4 

Choi T.Y. ( 2016 ). Wildlife on the roads . Seocheon: National Institute of Ecology.Seocheon: Nation- al Institute of Ecology .

5 

Clevenger A.P , Kociolek A.V. ( 2013 ). Potential impacts of highway median barriers on wildlife: State of the practice and gap analysis . Environmental Management, 52(5), 1299-1312 , doi:10.1007/s00267-0130155-0 .

6 

Collinson W.J. ( 2013 ). A standardised protocol for roadkill detection and the determinants of roadkill in the Greater Mapungubwe Transfrontier Conservation Area, Limpopo Province, South Africa (Doctoral dissertation) . Rhodes University, Eastern Cape, South Africa .

7 

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

8 

Ignatavicius G , Valskys V. ( 2018 ). The influence of time factors on the dynamics of roe deer collisions with vehicles . Landscape and Ecological Engineering, 14(2), 221-229 , doi:10.1007/s11355-017-0343-9 .

9 

Kim K.H , Choi H.H , Lee E.J , Park K.J. ( 2009 , 2009 ). A study on traffic accident reduction by improving expressway median barrier height . Proceedings of the 35th Korean Society of Civil Engineers . Hoengseong, Korea .

10 

Kim K , Serret H , Clauzel C , Andersen D , Jang Y. ( 2019 ). Spatio-temporal characteristics and predictions of the endangered leopard cat Prionailirus bengalensis euptilura road-kills in the Republic of Korea . Global Ecology and Conservation 19:e00673 , doi:10.1016/j.gecco.2019.e00673 .

11 

Kim K , Woo D.G , Seo H , Park T , Song E.G , Choi T.Y. ( 2019 ). Korea road-kill observation system: The first case to integrate road-kill data in national scale by government . Journal of Forest and Environmental Science, 35(4), 281-284 , doi:10.7747/JFES.2019.35.4.281 .

12 

Kioko J , Kiffner C , Phillips P , Patterson-Abrolat C , Collinson W , Katers S. ( 2015 ). Driver knowledge and attitudes on animal vehicle collisions in Northern Tanzania . Tropical Conservation Science, 8(2), 352366 , doi:10.1177/194008291500800206 .

13 

Malo J.E , Suárez F , Diez A. ( 2004 ). Can we mitigate animal-vehicle accidents using predictive models? Journal of Applied Ecology, 41(4), 701-710 , doi:10.1111/j.0021-8901.2004.00929.x .

14 

Park H , Woo D , Song E , Lim A , Lee B , Jang J., et al . ( 2018 ). Assessment of fence height to prevent roadkill of water deer (Hydropotes inermis) . Journal of Environmental Impact Assessment, 27(2), 232-239 , doi:10.14249/eia.2018.27.2.232 .

15 

Plante J , Jaeger J.A , Desrochers A. ( 2018 ). How do landscape context and fences influence roadkill locations of small and medium-sized mammals? Journal of Environmental Management, 235, 511-520 , doi:10.1016/j.jenvman.2018.10.093 .

16 

Rytwinski T , Soanes K , Jaeger J.A , Fahrig L , Findlay C.S , Houlahan J., et al . ( 2016 ). How effective is road mitigation at reducing road-kill? A meta-analysis . PLoS One, 11(11), e0166941 , doi:10.1371/journal.pone.0166941 .

17 

Saeki M , Macdonald D.W. ( 2004 ). The effects of traffic on the raccoon dog (Nyctereutes procyonoides viverrinus) and other mammals in Japan . Biological Conservation, 118(5), 559-571 , doi:10.1016/j.bio-con.2003.10.004 .

18 

Seo C , Thorne J.H , Choi T , Kwon H , Park C.H. ( 2015 ). Disentangling roadkill: the influence of landscape and season on cumulative vertebrate mortality in South Korea . Landscape and Ecological Engineering, 11(1), 87-99 , doi:10.1007/s11355-013-0239-2 .

19 

Smith-Patten B.D , Patten M.A. ( 2008 ). Diversity, seasonality, and context of mammalian roadkills in the southern Great Plains . Environmental Management, 41(6), 844-852 , doi:10.1007/s00267-008-9089-3 .

20 

Song E.G , Woo D.G , Kim K , Park T. ( 2020 ). Intensive survey on road-kill hotspots in South Korea . Seocheon:National Institute of Ecology.Seocheon:National Institute of Ecology .

21 

Song J.S , Lee K.J , Ki K.S , Jun I.Y. ( 2011 ). The Efficiency and improvement of the highway wild-life fences for decrease of mammals road-kill-In case of Manjong-Hongchun section on Jungang highway . Korean Journal of Environment and Ecology, 25(5), 649-657 .

22 

Williams S.T , Collinson W , Patterson-Abrolat C , Marneweck D.G , Swanepoel L.H. ( 2019 ). Using road patrol data to identify factors associated with carnivore roadkill counts . PeerJ, 7, e6650 , doi:10.7717/peerj.6650 .