Introduction

Wild boar (Sus scrofa) numbers have shown a continuous growth throughout Europe during the recent decades (Massei et al. 2014); hence, numerous studies have demonstrated increasing conflicts between wild boar populations and human activity, including research on wild boar-vehicle collisions (e.g. Balčiauskas and Balčiauskienė 2008; Lagos et al. 2012; Morelle et al. 2013; Rodríguez-Morales et al. 2013; Putzu et al. 2014). Although Estonia is situated close to the northern boundary of the species’ range, the wild boar population has experienced notable increase during recent decades (Estonian Environment Agency 2014) mainly due to supplementary feeding by hunters (Oja et al. 2014).

Wild boar incidents represent only 7 % of the total number of the country’s registered wildlife-vehicle collisions (WVCs) during the last decade; however, the increase of overall numbers of WBVCs in Estonia is alarming. In 1985–1990 on average, 79 WBVCs were registered each year (Mardiste 1992). In recent years, around 137 accidents with wild boar have been registered annually (Estonian Environmental Inspectorate, Estonian Police and Border Guard Board, unpublished data).

Therefore, proper knowledge about temporal and spatial patterns of WBVCs is essential to improve measures for reducing the collision numbers and collateral socio-economic and environmental damage. The aim of this study is to provide the first results of temporal patterns of wild boar-vehicle collisions for Estonia, using thoroughly reviewed data obtained from several authorities. We analysed seasonal, weekday and diurnal patterns of WBVCs.

Materials and methods

Study area

Our study area (45,227 km2) covers the whole Estonia, including islands. It is located between 57° 30′–59° 30′ N and 21°–28° E (Online Resource 1). The landscape is mainly flat with the mean altitude of 50 masl. Approximately 50 % of the country is covered with forests (dominant tree species Scots Pine (Pinus sylvestris) 32 %, Silver Birch (Betula pendula) 31 % and Norway Spruce (Picea abies) 17 %), 30 % is agricultural land, 6 % internal waters and 5 % bogs (Estonian Environment Agency 2014). According to the 2011 census, Estonia has an average population density of 30 people per km2 but the distribution is very uneven. Majority of the population (43 %) is concentrated to the capital Tallinn and surrounding Harju County whereas 55 % of Estonian territory is uninhabited (Statistics Estonia 2015). Road network consists of national (28 %), local (41 %) and private and forest roads (31 %) with total length of 58 787 km and an average road density of 1.3 km/km2 (Estonian Road Administration 2014). National roads are divided to main (10 %, 1607 km), basic (15 %, 2406 km) and secondary (75 %, 12 476 km) roads. The speed limit on the national roads is usually 90 km/h all year round and 110 km/h in summer on some short four-lane road sections. Average traffic density was 812 vehicles per day on all the national roads and 4388 vehicles per day on the main roads in 2013 (Kaal et al. 2014). The number of motor vehicles in Estonia has increased from 571,000 in 2004 to 750,000 in 2013 (Statistics Estonia 2015).

Collision data

We used three available data sets from the period of 2004–2013 in our study: (i) the list of environmental emergency hotline calls—the information received mostly from road users but also from hunters and police officers is gathered by the Environmental Inspectorate (EI). According to law, a driver who hits a big game animal species has to report to the EI or local hunters; (ii) data about WVCs resulting in human injuries collected by the police; (iii) data about WVCs resulting in property damage collected by the police and insurance companies. In this instance, only the cases when the vehicle had comprehensive insurance were notified, as the regular obligatory car insurance does not cover damage caused by wild animals. Accidents involving other species than wild boar were excluded from the analysis. Remaining entries were thoroughly reviewed. Duplications (the same WBVC was registered by both, the EI and the police) were removed. On the basis of the description of the case, reports were divided into three groups: (i) A—accurate (the accident had happened immediately prior to the notification), (ii) N—not sure (the accident could have happened immediately prior to the notification or slightly earlier), and (iii) I—inaccurate (the accident had happened earlier; for example, someone noticed an animal carcass in a road ditch). Only ‘accurate’ and ‘not sure’ entries were selected for further analysis, resulting in a data set of 918 WBVC reports.

Data analysis

Firstly, we analysed seasonal and weekday patterns of WBVCs. We set the hypothesis that the distribution of collisions would be different from the expected average for each of these timescales. The chi-square goodness-of-fit test was used to check the hypothesis (van Emden 2008). Statistical analysis was carried out using R version 3.1.2 (R Core Team 2014).

Secondly, diurnal patterns of WBVCs were examined. Before the analysis, all WBVC times were converted into Eastern European Time (UTC + 2 h), hereafter referred to as local time (LT). This step was necessary as summer incidents were originally registered in daylight saving time.

For all WBVCs, the following astronomical parameters were computed: time in seconds to sunrise and after sunset, altitude of the sun and the moon in degrees. Ephem, an astronomy module of Python programming language, was used for the calculations. A special script was developed on the basis of Ephem. The script inputs geographical coordinates and times of WBVCs and returns their astronomical parameters. The exact geographical coordinates of collisions are not registered; thus, the coordinates of the approximate centres of Estonian counties were used as accident locations. As Estonian counties are typically much less than 100 km across, the sunrise and sunset times do not differ by more than a few minutes within the counties. Similarly, the altitude of the sun and the moon can vary by no more than one degree. Thus, the use of the county centres is accurate enough considering that the data of the WBVCs is not free of a small timing errors in a range of a few minutes. The traditional definition of twilight is used in the present study. According to the definition, twilight is the time period when the altitude of the sun is between 0 and −18°. There are three subcategories of twilight: civil twilight (brightest, the sun is 0–6° below the horizon), nautical twilight (the sun is 6–12° below the horizon) and astronomical twilight (darkest, the sun is 12–18° below the horizon).

Results and discussion

The variation in numbers of registered WBVCs was noticeable during 2004–2013 (Fig. 1). The lowest number of collisions was registered in 2004 (16) and the largest number in 2013 (174).

Fig. 1
figure 1

Wild boar harvest and collision numbers in Estonia, 2004–2013 (Estonian Environment Agency 2014; Estonian Environmental Inspectorate, Estonian Police and Border Guard Board)

Full size image

Supposedly, the dramatic increase in accident reports during 2004–2007 was due to several reasons. As there was no wildlife-vehicle collision reporting system in Estonia before 2000, one reason was probably rising awareness of road users, who started to report the collisions more often. Secondly, the occurrence of ungulate collisions increases with higher traffic volume but this relationship is not linear (Romin and Bissonette 1996; Zuberogoitia et al. 2014; Valero et al. 2015). According to Thurfjell et al. (2015) the collision risk related to wild boar is highest at intermediate traffic levels. Average traffic density on main roads increased from 3520 vehicles per day in 2004 to 4740 vehicles per day in 2007.

Furthermore, the number of wild boar has risen from 17,000 in 2004 to 22,320 in 2013 with a peak of 23,450 in 2009 (Estonian Environment Agency 2014). This is likely a result of extensive supplementary feeding primarily in the wintertime (Oja et al. 2014) and avoidance of hunting adult females, which both have been common management practise in Estonia during the 2000s (Veeroja and Männil 2014). Such high density of wild boar population has also been associated to proliferation of African swine fever virus (ASFV). As several occasions of ASFV were diagnosed in Latvia, Lithuania and Poland earlier in 2014, the first outbreak in Estonia, near Latvian border, was recorded in September 2014. Supposedly, the infection spread to South Estonia from Latvia by free-roaming wild boar (Nurmoja 2016; Oļševskis et al. 2016). However, outbreaks in northeastern part of the country have been linked to human activity—through infected meat and food waste (Nurmoja 2016).

In monthly distribution, we noticed statistically significant deviation from the expected average. WBVCs (χ 2 = 455.7, df = 11, p < 0.001, n = 918) peak in September–December with the highest frequency in October and are the least frequent in February, April and May (Fig. 2a). This pattern is similar to Lithuania (Balčiauskas and Balčiauskienė 2008), Wallonia in southern Belgium (Morelle et al. 2013) and also Italy (Putzu et al. 2014). In Galicia (Spain) and southern Sweden, the peak occurs 1 month later from October to January (Lagos et al. 2012; Rodríguez-Morales et al. 2013; Thurfjell et al. 2015). As suggested by Thurfjell et al. (2015), wild boar-vehicle collisions happen when traffic is at intermediate levels and female wild boars are active. These two factors probably meet during the long and dark late fall/winter nights. Poor light and road surface conditions (e.g. rain, snow, ice) may also increase the collision risk (Neumann et al. 2012).

Fig. 2
figure 2

a Annual course of diurnal distribution of WBVCs with a daily step. Individual WBVCs are represented as dots and darker background shading indicates higher frequency of WBVCs. Thick lines show the sunrise and sunset times. Civil, nautical and astronomical twilight times are bounded by thin, dashed and dotted lines, respectively. b Weekday distribution of WBVCs

Full size image

The WBVCs were not uniformly distributed between days of the week (Fig. 2b). More incidents occur at weekends with a notable peak on Friday, and the difference from the expected average is statistically significant (χ 2 = 17.0, df = 6, p < 0.01, n = 918). The lowest collision risk was observed on Thursday. High collision risk at weekends is presumably related to traffic volume as no biological, ecological or behavioural factors depending on weekdays are known (Steiner et al. 2014; Hothorn et al. 2015). We did not manage to obtain detailed traffic data for all roads but traffic count studies carried out in 2006 and 2007 on one of the main roads (E263 Tallinn-Tartu-Võru-Luhamaa) suggested that traffic volume increases sharply on Friday afternoon (Koonik 2008; see also Järv et al. 2012). This is in compliance with other studies that relate the weekend ungulate-vehicle collision peaks to increased traffic (Dussault et al. 2006) and the drivers’ habits (Lagos et al. 2012). In Estonia, numerous city dwellers drive to their country houses for the weekend. On Friday, people also drive to the parties at night and return early in the morning like discussed by Lagos et al. (2012).

Regarding diurnal patterns, the frequency of wild boar collisions increases suddenly if the sun is more than 6–7° below the horizon and remains high even in the complete darkness when the sun is more than 18° below the horizon (Fig. 2a). As mostly the main activity of ungulates is related to sunrise and sunset, wild boars are generally known as nocturnal animals (Boitani et al. 1994; Thurfjell et al. 2015). This is in line with our finding that the majority of collisions related to this species occurred in nighttime.

Conclusion

Collisions with large free-ranging animals are an increasing problem all over the world. Significant temporal variations in wildlife collisions occur in different regions (Groot Bruinderink and Hazebroek 1996). Thus, it is important to identify the local patterns. Both spatial and temporal data are necessary for choosing suitable mitigation measures. We investigated seasonal, weekday and diurnal patterns of traffic accidents related to wild boar. Accurate temporal data enables to warn road users about periods of high collision risk by using temporary electronic warning signs, mobile apps or road safety campaigns, for example.