key: cord-0032326-4wjttlcs
authors: Balčiauskas, Linas; Stratford, Jos; Kučas, Andrius; Balčiauskienė, Laima
title: Factors Affecting Roadkills of Semi-Aquatic Mammals
date: 2022-05-13
journal: Biology (Basel)
DOI: 10.3390/biology11050748
sha: a0b192d9d156d6544f381a44cf05ca415dc68c83
doc_id: 32326
cord_uid: 4wjttlcs

SIMPLE SUMMARY: Semiaquatic mammal roadkill, specifically the European beaver, Eurasian otter, American mink and muskrat was analyzed in Lithuania in the period 2002–2021. From almost 40,000 recorded roadkill cases for all species, these species were registered 60, 22, 26 and 3 times, respectively, thus being rare events. Registered beaver roadkill numbers correlated with an increased registration effort and traffic intensity, while otter roadkill correlated with a registration effort only, and mink roadkill correlated with an increased population (estimated through its proxy and the number of hunted individuals per year). The roadkill was not always in close proximity to waterbodies, with 38–54% of roadkill of these species occurring over 200 m from the nearest water source. With the American mink and muskrat being invasive species in the EU and the Eurasian otter protected in many countries, it could be considered valuable to enhance the registrations of their roadkill (using targeted efforts by drivers, hunters or other citizen scientists) in order to obtain the extrapolated amount of roadkill and to use this knowledge in species management. ABSTRACT: We previously showed that registration efforts and traffic intensity explain 90% of variation in mammal roadkill numbers, 70% of variation in the numbers of recorded species and 40% of diversity variation. Here we analyze semiaquatic mammal roadkill in Lithuania in 2002–2021, relating these to the monitoring effort. From 39,936 analyzed roadkill, the European beaver (Castor fiber) was registered 60 times, American mink (Neovison vison) 26 times, otter (Lutra lutra) 22 times and muskrat (Ondatra zibethica) 3 times. The average roadkill indexes were 0.000065, 0.00076, 0.00061 and 0.00010 ind./km/day, and the extrapolated annual roadkill for the country was 44–357, 36–456, 49–464 and 89–144 individuals, respectively. Beaver roadkill numbers correlated with the registration effort and traffic intensity, otter roadkill with registration effort only and mink with hunting bag (number of hunted individuals per year). Roadkill was not always related to proximity to water, with 38–54% of roadkill occurring over 200 m from the nearest water source. With American mink and muskrat being invasive species in the EU and otter protected in many countries, it is valuable to enhance the registrations of their roadkill (using targeted efforts by drivers, hunters or other citizen scientists) to obtain the extrapolated amount of roadkill and to use this knowledge in species management.

Semiaquatic mammals form an ecological group of animals whose lifestyles are related both to aquatic and terrestrial habitats with varying degrees of dependence [1, 2] . Four such medium-sized species are present in Lithuania, namely European beaver (Castor fiber), Eurasian otter (Lutra lutra), American mink (Neovison vison) and muskrat (Ondatra zibethicus) [3] , the list of species being similar to that in North America [4] . Of these species, C. fiber is considered an ecosystem engineer as it modifies ecosystems and enhances habitat suitability for the other animals [5, 6] .

The habitats used by this group of mammals include various waterbodies, such as swamps, streams, rivers and lakes [7] with seasonal water level changes having a significant

We used data collected by professionals at the Nature Research Centre. A registration session was defined as registering roadkill along any length of a same-numbered road travelled on the same day. A total of 3570 such sessions were conducted nationally from 2007-2021, with the total length of all driven routes equaling 275,300 km. The most frequent registration sessions were performed on the main roads A1, A2, A14 and A8 (405, 405, 400 and 210 sessions, respectively). The number of sessions on other main, national and regional roads (Figure 1 ) varied from 1-125. Data collected by the Lithuanian Police Traffic Supervision Service between 2002 and 2021 were also used. In total, we analyzed data on 39,936 WVCs involving wild mammals (MVCs). In both datasets, individuals not identifying with the species level (including "mustelid", "small carnivore", "hare", "big game", etc.) were pooled under category "unknown". zibethicus. We tested (i) if semiaquatic mammal roadkill was related to proximity to waterbodies (rivers and lakes) or, in the cases of C. fiber and O. zibethicus, to wetlands and (ii) if there is any relationship among roadkill frequencies and population/hunting bag sizes, transport intensity and sampling effort.

We used data collected by professionals at the Nature Research Centre. A registration session was defined as registering roadkill along any length of a same-numbered road travelled on the same day. A total of 3570 such sessions were conducted nationally from 2007-2021, with the total length of all driven routes equaling 275,300 km. The most frequent registration sessions were performed on the main roads A1, A2, A14 and A8 (405, 405, 400 and 210 sessions, respectively). The number of sessions on other main, national and regional roads ( Figure 1 ) varied from 1-125. Data collected by the Lithuanian Police Traffic Supervision Service between 2002 and 2021 were also used. In total, we analyzed data on 39,936 WVCs involving wild mammals (MVCs). In both datasets, individuals not identifying with the species level (including "mustelid", "small carnivore", "hare", "big game", etc.) were pooled under category "unknown".

Domestic mammal roadkill (n = 3150) was not analyzed in detail, as these species are mostly related to settlements not habitats. The most frequent roadkill in this category was dogs (n = 2134), cattle (n = 414), horses (n = 220) and cats (n = 220), followed by sheep, goats, pigs and rabbits in descending numbers. Domestic mammal roadkill (n = 3150) was not analyzed in detail, as these species are mostly related to settlements not habitats. The most frequent roadkill in this category was dogs (n = 2134), cattle (n = 414), horses (n = 220) and cats (n = 220), followed by sheep, goats, pigs and rabbits in descending numbers.

In 2021, there were 21,237 km of state roads of national significance, these being 1751 km of main roads (AADT 3000-20,000), 4928 km of national roads (AADT 500-3000) and 14,559 km of regional roads (AADT up to 500) [41, 42] . Changes in road lengths since 2017 [43] have been negligible and, thus, were ignored.

Roadkill indices were calculated as the number of killed animals per km/day. "Empty routes", where roadkill had already possibly been removed, were included into the yearly index. Estimations of the number of animals killed at the country level for each of the four species were based on the length of the season of actual registrations ( Figure 2 ). L. lutra and N. vison were not registered in January and February; therefore, the season length was set at 300 days, while C. fiber was not registered in February, and thus the season length was set at 330 days. Based on the ecology of species [44] and the registered roadkill, the season length for O. zibethicus was set at 240 days. Calculations were done using the total lengths of the main and national roads only (6679 km) as the roadkill registration efforts on the regional roads were not sufficient. The presented estimations can, therefore, be seen as minimum numbers. In 2021, there were 21,237 km of state roads of national significance, these being 1751 km of main roads (AADT 3000-20,000), 4928 km of national roads (AADT 500-3000) and 14,559 km of regional roads (AADT up to 500) [41, 42] . Changes in road lengths since 2017 [43] have been negligible and, thus, were ignored.

Roadkill indices were calculated as the number of killed animals per km/day. "Empty routes", where roadkill had already possibly been removed, were included into the yearly index. Estimations of the number of animals killed at the country level for each of the four species were based on the length of the season of actual registrations ( Figure 2 ). L. lutra and N. vison were not registered in January and February; therefore, the season length was set at 300 days, while C. fiber was not registered in February, and thus the season length was set at 330 days. Based on the ecology of species [44] and the registered roadkill, the season length for O. zibethicus was set at 240 days. Calculations were done using the total lengths of the main and national roads only (6679 km) as the roadkill registration efforts on the regional roads were not sufficient. The presented estimations can, therefore, be seen as minimum numbers. Table 1 . This information was collected over a longer time period, but earlier information is not available as open data anymore [45] . N. vison and O. zibethicus are not included in game survey programs; therefore, only information on the bag size was available. L. lutra is a protected species; therefore, there is no hunting bag, and the species is not included in Table 1 . As a result, the numbers of L. lutra and N. vison that we used are based on monitoring data and the distribution of these species along rivers [12, [46] [47] [48] .

In summary, the population of C. fiber in the years since 2010 has ranged from 40 to 50 thousand individuals (Table 1), while that of L. lutra has been 3000-5000 individuals and that of N. vison in the region of 5000-10,000 individuals [48, 49] . The numbers of O. zibethicus have sharply declined since the 1990s and based on the reduced hunting bag, could be 2000-3000 in the last decade [48] . Table 1 . This information was collected over a longer time period, but earlier information is not available as open data anymore [45] . N. vison and O. zibethicus are not included in game survey programs; therefore, only information on the bag size was available. L. lutra is a protected species; therefore, there is no hunting bag, and the species is not included in Table 1 . As a result, the numbers of L. lutra and N. vison that we used are based on monitoring data and the distribution of these species along rivers [12, [46] [47] [48] . In summary, the population of C. fiber in the years since 2010 has ranged from 40 to 50 thousand individuals (Table 1), while that of L. lutra has been 3000-5000 individuals and that of N. vison in the region of 5000-10,000 individuals [48, 49] . The numbers of O. zibethicus have sharply declined since the 1990s and based on the reduced hunting bag, could be 2000-3000 in the last decade [48] .

Data on annual average daily traffic (AADT) for the roads with semiaquatic mammal roadkill were obtained from the Lithuanian Road Administration under the Ministry of Transport and Communications [42] .

Distances from the roadkill to the nearest waterbody were investigated at the (1) medium spatial scale using the Corine land cover inventory (CLC) 1:50,000 [50] and (2) fine scale using the spatial dataset of (geo) reference base cadastre (GRPK) 1:10,000 (https://www. geoportal.lt/metadata-catalog/catalog/search/resource/details.page?uuid=%7B513C0C2 9-0447-CB3D-4585-2390144D20D2%7D accessed on 10 February 2022). In the CLC, two land cover classes (411 inland marshes and 412 peat bogs) were used as "wetlands" and two classes of waterbodies (511 water courses and 512 waterbodies) as "water". Due to the absence of roadkill in the relevant areas, two classes (521 coastal lagoons and 523 sea and ocean) were not analyzed. In the GPRK, we used two habitats, namely "rivers" and "lakes". Distances from the roadkill points to the closest polygon were measured using near distance measurement tools (https://pro.arcgis.com/en/pro-app/2.8/tool-reference/ analysis/near.htm accessed on 5 February 2022) that are part of the standard ArcGIS Proximity toolset (https://pro.arcgis.com/en/pro-app/2.8/tool-reference/analysis/anoverview-of-the-proximity-toolset.htm accessed on 5 February 2022). As distance statistics, species-based averages and SE were calculated. After finding that the distribution of distances did not conform to normal, between-species differences were evaluated with the nonparametric Kruskal-Wallis H test.

At the country scale, we applied a multiple regression analysis to the annual C. fiber and N. vison roadkill to test the possible cumulative influence of the categorical predictors, specifically annual population and hunting bag sizes, sampling effort and AADT. The confidence level was set as p < 0.05. At the local scale, we tested if there was a correlation among roadkill, sampling effort and AADT using the road as the sampling unit. The Pearson correlation coefficient was used. Calculations were done in Statistica for Windows, ver. 6.0 (StatSoft, Inc., Tulsa, OK, USA).

The proportion of semiaquatic mammal roadkill and the 95% CI for species proportion among all wild mammal roadkill were calculated with the Wilson method of the score interval [51] using OpenEpi epidemiological software [52] . Differences in the proportions of roadkill were evaluated using the G test online calculator [53] .

In the analyzed roadkill, 32 wild mammal species were registered in 2002-2021 (Table 2 ). Unidentified species accounted for 8.30%, with the pooled number for roadkill being < 20 per species -0.12% (Table 2) . Semiaquatic mammals accounted for only a marginal proportion of the total amount of roadkill registered, this being 0.15% of all roadkill (CI = 0.12-0.19%) for C. fiber, 0.065% (CI = 0.044-0.095%) for N. vison, 0.055% (CI = 0.036-0.083%) for L. lutra and 0.0075% for O. zibethicus. Based on the G-test (G = 62.43), the proportion of C. fiber roadkills was higher (p < 0.001) and that of O. zibethicus lower (p < 0.003) than the other two species. 

The spatial distributions of the semiaquatic roadkill species show C. fiber registrations to be scattered across the county (Figure 3a the north of the country (Figure 3b ). Most N. vison roadkill was concentrated in the eastern part of the country (Figure 3c) , especially on the main road A14 (Figure 1 ), this characterized by the highest sampling effort. the north of the country (Figure 3b ). Most N. vison roadkill was concentrated in the eastern part of the country (Figure 3c) , especially on the main road A14 (Figure 1) , this characterized by the highest sampling effort. 

At the local scale, using the road as the sampling unit, we analyzed correlations between semiaquatic animal roadkill numbers and the sampling effort ( Figure 6 Traffic intensity was correlated with C. fiber roadkill numbers (r = 0.68, p < 0.001), but did not correlate with L. lutra or N. vison roadkill numbers (r = 0.34 and r = −0.07, both NS; Figure 7 ). However, the semiaquatic mammals did tend to be killed on the main roads 

At the local scale, using the road as the sampling unit, we analyzed correlations between semiaquatic animal roadkill numbers and the sampling effort ( Figure 6 ) and traffic intensity (Figure 7) . The enhanced registration effort yielded higher numbers of registered roadkill of C. fiber and L. lutra (r = 0.9, p < 0.001 and r = 0.69, p < 0.02), but not N. vison 

Due to inaccuracies arising from the limited amount of survey, we exclude the predicted amount of L. lutra roadkill in 2004. The figure calculated was in the order of 6000 individuals per year, a strikingly high figure. 

At the local scale, using the road as the sampling unit, we analyzed correlations between semiaquatic animal roadkill numbers and the sampling effort ( Figure 6 ) and traffic intensity (Figure 7) . The enhanced registration effort yielded higher numbers of registered roadkill of C. fiber and L. lutra (r = 0.9, p < 0.001 and r = 0.69, p < 0.02), but not N. vison (r = 0.53, NS). Most roadkill was registered on a relatively short segment of the A14 main road ( Traffic intensity was correlated with C. fiber roadkill numbers (r = 0.68, p < 0.001), but did not correlate with L. lutra or N. vison roadkill numbers (r = 0.34 and r = −0.07, both NS; Figure 7 ). However, the semiaquatic mammals did tend to be killed on the main roads Traffic intensity was correlated with C. fiber roadkill numbers (r = 0.68, p < 0.001), but did not correlate with L. lutra or N. vison roadkill numbers (r = 0.34 and r = −0.07, both NS; Figure 7 ). However, the semiaquatic mammals did tend to be killed on the main roads that were characterized by the highest AADT and traffic speed. Such roads accounted for 100% of N. vison and O. zibethicus roadkill, 77.3% of L. lutra and 46.7% of C. fiber. National roads, with medium AADT, accounted for 35.0% of C. fiber roadkill and 18.2% of L. lutra roadkill, while on regional roads with low AADT, the respective values were just 10.0% and 4.5%.

At the country scale, we used regression analysis to see if the amount of annually registered semiaquatic animal roadkill was related to the annual averages of AADT, the sum of sampling effort, the hunting bag size of C. fiber and N. vison and the population size of C. fiber. The model for C. fiber was not significant (F 3,14 = 1.97, p = 0.16) and explained less than 20% of roadkill variation (R 2 = 0.19. Of the four analyzed factors for this species, only AADT was significant (F = 2.39, p < 0.05). Population number, hunting bag size and sampling effort did not correlate with the amount of roadkill (r = 0.23, 0.19 and 0.28 respectively, all NS; Figure 8 ). At the country scale, we used regression analysis to see if the amount of annually registered semiaquatic animal roadkill was related to the annual averages of AADT, the sum of sampling effort, the hunting bag size of C. fiber and N. vison and the population size of C. fiber. The model for C. fiber was not significant (F3,14 = 1.97, p = 0.16) and explained less than 20% of roadkill variation (R 2 = 0.19. Of the four analyzed factors for this species, only AADT was significant (F = 2.39, p < 0.05). Population number, hunting bag size and sampling effort did not correlate with the amount of roadkill (r = 0.23, 0.19 and 0.28 respectively, all NS; Figure 8 ). The model for N. vison was significant (F3,15 = 21.3, p < 0.0001) and explained nearly 80% of roadkill variation (R 2 = 0.77). Both the size of the hunting bag and sampling intensity were significant (F = 5.79 and F = 4.47, both p < 0.001), while AADT was not (F = 0.43, NS). Numbers of hunted N. vison correlated with the roadkill of the same year strongly and positively (r = 0.75, p < 0.001; Figure 8 ) and so did sampling intensity (r = 0.61, p < 0.001). L. lutra and N. vison, although about 10 and 5 times less numerous in terms of population than C. fiber, had relatively high levels of roadkill, these still amounting to about half of the C. fiber roadkill numbers. In O. zibethicus, the amount of registered roadkill was too small to make any comparisons. The model for N. vison was significant (F 3,15 = 21.3, p < 0.0001) and explained nearly 80% of roadkill variation (R 2 = 0.77). Both the size of the hunting bag and sampling intensity were significant (F = 5.79 and F = 4.47, both p < 0.001), while AADT was not (F = 0.43, NS). Numbers of hunted N. vison correlated with the roadkill of the same year strongly and positively (r = 0.75, p < 0.001; Figure 8 ) and so did sampling intensity (r = 0.61, p < 0.001). L. lutra and N. vison, although about 10 and 5 times less numerous in terms of population than C. fiber, had relatively high levels of roadkill, these still amounting to about half of the C. fiber roadkill numbers. In O. zibethicus, the amount of registered roadkill was too small to make any comparisons.

We found the differences in the distances from the roadkill to the nearest water source to not be significant among semiaquatic mammal species, regardless of the scale (size) of the water source: CLC water, Kruskal-Wallis H 3,92 = 1.37, p = 0.71; CLC wetland, H = 2.87, p = 0.41; GRPK river, H = 3.22, p = 0.36 and GRPK lake, H = 0.94, p = 0.81. Measured at the highest possible map scale, 26% of C. fiber roadkill occurred less than 50 m from the smallest possible water source, 43% less than 100 m and 62% less than 200 m from water. For L. lutra, the respective figures were 24%, 38% and 48%, while for N. vison they were 15%, 19% and 46%. In other words, 38-54% of roadkill in these species occurred over 200 m from the nearest water ( Figure 9 ). Differences among species were not significant. Distances to large water sources, including wetlands, were much bigger. The average distances are presented in Table 3 . 

Upon completing our analysis of roadkill of the four affected semiaquatic mammal species in Lithuania, we found roadkill to be a rare event. The proportion of C. fiber was 0.15% of all roadkill on average, with the proportions of N. vison and L. lutra Otter roadkill is of particular interest as this species was decreasing for a long time in many western countries. In Britain, L. lutra roadkill amounted to 673 animals in 1971-1996 [54] , these numbers increasing later, with 500 records in the southwest of the country alone between 1984 and 2004 [55] . In a 3000 sq. km area of northeastern Germany, there were 88 instances of L. lutra roadkill in 1990-2003 [56] , while the total amount of L. lutra roadkill in eastern Germany was 746 ind. In the period 1957-1998 [57] . In the Czech Republic, there were 316 instances of L. lutra roadkill from 1990 to 2011 [58] , while 53 ind. Were recorded in Hungary during 2006-2011 [59] . These relatively low numbers are in accordance with those found in Denmark [60] , but 109 instances of otter roadkill in northern Israel during 1965-2004 represented an estimated 5% loss of the population each year [61] . Based on our data, the L. lutra roadkill numbers in Lithuania account for between 1% and 10% of the population annually, with data from the last three decades not showing any overall population decrease [12, 46, 48] .

Higher traffic intensity is mentioned as a factor for increased L. lutra roadkill frequency [58] , though our study did not confirm this. At present, neither spatial nor temporal correlations allow us to determine whether or not the number of otter casualties has had an impact on otter populations [55] .

In general, the global trend of an increase in roadkill numbers over time is accepted, see references in [15, 62] , though during the COVID-19 lockdowns, there were decreases in roadkill, at least in some species [63] . In this context, the numbers of semiaquatic animal roadkill in various countries are very low. For various mammal species, low roadkill rates may be explained as due to the species being a habitat specialist [64] . In terms of semiaquatic species, it may be that aquatic and wetland habitats are not frequently around roads.

In Poland, a neighboring country of Lithuania, the proportion of O. zibethicus was 0.50% of total mammal roadkill in 2000-2016, while that of C. fiber and of N. vison was both 0.38% [65] . In the southwestern part of Poland in the period 2001-2003, no C. fiber roadkill was registered, while that of O. zibethicus was 0.52% and N. vison 0.26% out of 383 wild mammals [66] . In northeastern Poland, on the road crossing the valley of the large river Biebrza, O. zibethicus and N. vison roadkill each accounted for 1.25% out of 80 wild mammal registrations [67] .

No semiaquatic mammal roadkill was registered in Croatia between 2005 and 2008 [68] or between 2007 and 2009 [69] . The same absence of semiaquatic mammal roadkill was characteristic to a study of 1175 instances of roadkill in Brittany, western France [70] , 40,000 instance of roadkill in Czech Republic [27] and ca. 2500 roadkill registrations in Walonia, Belgium [71] .

However, there were also cases with much higher proportions of semiaquatic animal roadkill. In the Republic of Ireland in 2008-2010, out of 548 roadkill instances, 15 instances were N. vison and 10 L. lutra, comprising 2.74% and 1.82%, respectively [72] . In areas where roads run in river valleys and wetlands, the proportion of O. zibethicus roadkill may reach much higher values. There are two investigations from North America that report very high proportions of O. zibethicus roadkill-10.4% in NY State [73] and 15.75% in California-in an area where a highway traverses a tidal marsh and diked baylands [74] .

Our data show that 38-54% of semiaquatic mammal roadkill occurred more than 200 m from the nearest water. In this, we included the smallest possible water sources, such as land reclamation ditches, which are nearly or completely dry in the warm season. For L. lutra, such a phenomenon was known before. In Germany, between 1985 and 1993, 46.84% of otter roadkill was registered at sites with no adjacent waterbodies [75] , and 32% were further than 200 m [56] . By contrast, 67% of all L. lutra casualty records occurred in a 100 m wide zone surrounding fresh water and the coast in Great Britain [54] and 57% in Israel [61] . Data about the other species have not been published before.

Of the other peculiarities of semiaquatic animal roadkill, it should be mentioned that these events are not representative of the population [76] ; there are behavior differences in relation to oncoming vehicles that may lead to differences in their roadkill [77] , and possible mitigations can create barrier effects for them [54, 73] . It was shown that the survival of the American beaver (Castor canadensis) is inverse to habitat suitability; thus, a new approach might be required to understand semiaquatic mammal distribution and movements [78] .

In Europe, N. vison densities vary from below one to over ten individuals per 10 km of water length [79] . Being a semiaquatic, generalist predator, N. vison can live in various habitats: in Greece, two thirds of records were near water, but many animals were also recorded far from water [80] . In central Italy, roadkill was the first evidence of species spread [81] . Avoiding urbanized areas and isolated ponds in the agricultural landscape is characteristic to the species [82] . During the dry season and droughts, N. vison readily uses terrestrial habitats, though the species is subjected to increased mortality when away from streams [83] . Under conditions of climate change, the drying of landscapes will affect local hydrology regimes, changing semiaquatic mammal population dynamics and movements [84] ; therefore, the amount of roadkill registered not in proximity to waterbodies may further increase. This may apply to O. zibethicus as it is the semiaquatic mammal species most restricted to aquatic habitats [83] . Finally, habitat features may be not the main factors affecting occupancy by semiaquatic mammals, at least for the river otter (Lontra canadensis) and N. vison [85] .

The fate of introduced and reintroduced semiaquatic species in Lithuania varied. Following the release from captivity in 1950 and the introduction in 1953, N. vison numbers continuously increased [10] , totally suppressing the native European mink (Mustela lutreola) [11, 14] .

After the re-establishment and reintroduction of C. fiber to Lithuania in the 1940s and 1950s, they rapidly colonized the entire country [9, 10] . According to some expert opinions, their numbers have been close to 100,000 individuals in recent decades [10] , i.e., twice the official estimate (see Table 2 ). With the collapse of the fur market, these numbers are near the carrying capacity of the habitats.

O. zibethicus was introduced into eastern and southern parts of Lithuania in 1954, where it still remains most abundant. It is scarce in the north of the country [10, 11] . Across the country, however, the species declined by about 94% in the period from 1989 to 2013 [86, 87] . The success of the initial introduction is considered to be related to similarities to the climate and habitats within the native species range, as well as to the presence of a free ecological niche [86] . By contrast, the reasons for the decline are not well understood, and the presence of a high number of native predators may be involved, as may pathogens [87] .

As N. vison and O. zibethicus are considered invasive species in the EU, while L. lutra is protected in many countries of Europe, it could be considered valuable to enhance registrations of their roadkill (using targeted efforts by drivers, hunters or other citizen scientists) in order to obtain the extrapolated amount of roadkill and to use this knowledge in species management.

Semiaquatic mammal roadkill is rare, even in C. fiber accounting for only 0.15% of total roadkill numbers. The national annual roadkill numbers were 44-357 in the case of C. fiber, 36-456 in L. lutra, 49- Institutional Review Board Statement: Not applicable.

Data Availability Statement: Due to ongoing investigation, data of this study are available from the corresponding author upon personal request.

Substrate utilization and feeding strategies of mammals: Description and classification

Semi-Aquatic Mammals: Ecology and Biology

Spatial relations among semi-aquatic mammals on the riverside

Accessing and mobilizing "new" data to evaluate emerging environmental impacts on semi-aquatic mammals

Organisms as ecosystem engineers

The beaver facilitates species richness and abundance of terrestrial and semi-aquatic mammals

Aquatic habitat use in a semi-aquatic mammal: The Eurasian beaver

Occupancy of Semi-Aquatic Mammals in an Urban Landscape

Habitat use and selectivity by beavers (Castor fiber) in anthropogenic landscape

Atlas of Lithuanian Mammals, Amphibians and Reptiles

Otter distribution in Lithuania in 2008 and changes in the last decade

Changes in the territory use by semi-aquatic mammals in some of rivers in Lithuania

Red Data Book of Lithuania. Animals, Plants, Fungi; Rašomavičius

Spatiotemporal Analysis of Road Crashes with Animals in Poland

Importance of professional roadkill data in assessing diversity of mammal roadkills

The small, the forgotten and the dead: Highway impact on vertebrates and its implications for mitigation strategies

Diversity of vertebrates affected by roadkill and areas of greatest impact in Ruta III General Elizardo Aquino

Use of unfenced highway underpasses by lowland tapirs and other medium and large mammals in central-western Brazil

Roadkill of medium to large mammals along a Brazilian road (BR-262) in Southeastern Brazil: Spatial distribution and seasonal variation

Use of highway culverts by the water opossum (Chironectes minimus) in southeastern Brazil

Long-term monitoring by roadkill counts of mammal populations living in intensively cultivated landscapes

The value of monitoring wildlife roadkill

Impact of Road Fencing on Ungulate-Vehicle Collisions and Hotspot Patterns

Spatiotemporal Analysis of Vehicle Collisions Involving Wild Boar and Roe Deer in NW Spain

Vertebrate Road-Kill Patterns in Mediterranean Habitats: Who, When and Where

Cz: A System for Evidence of Animal-Vehicle Collisions along Transportation Networks

Trends and Characteristics of Animal-Vehicle Collisions in the United States

Predicting Spatiotemporal Patterns of Road Mortality for Medium-Large Mammals

Wildlife-Vehicle Collisions-Influencing Factors, Data Collection and Research Methods

The Effect of Traffic Intensity and Animal Activity on Probability of Ungulate-Vehicle Collisions in the Czech Republic

Animal-Related Motorcycle Collisions in North Dakota

Protecting the protected: Reducing wildlife roadkill in protected areas

Conservation threats from roadkill in the global road network

Spatial Patterns and Factors Influencing Small Vertebrate Fauna Road-Kill Aggregations

Roadside Vegetation Influences Clustering of Ungulate Vehicle Collisions

Roadkill-Data-Based Identification and Ranking of Mammal Habitats

Roadkills as a Method to Monitor Raccoon Dog Populations

Ungulate: Vehicle Collision Rates Are Associated with the Phase of the Moon

Temporal Patterns of Ungulate-Vehicle Collisions in Lithuania

VĮ Lietuvos Automobilių Kelių Direkcija. Eismo Apskaita Valstybinės Reikšmės Keliuose

Analysis and Assessment of Lithuanian Road Accidents by AHP Method

Ondatra zibethicus) decline after the expansion of American mink (Neovison vison) in Poland

Semi-aquatic mammal environment correlates in South Lithuanian river valleys

Ecology of semi-aquatic mammals in some landscapes of Lithuania: Abundance, community structure and distribution

Interval Estimation for a proportion

Open Source Epidemiologic Statistics for Public Health

Patterns of otter Lutra lutra road mortality in Britain

Otter road casualties

Patterns of otter Lutra lutra road mortality in a landscape abundant in lakes

Mortality patterns of otters (Lutra lutra) from eastern Germany

Causes of deaths of Lutra lutra in the Czech Republic (Carnivora: Mustelidae)

Interchange as the main factor determining wildlife-vehicle collision hotspots on the fenced highways: Spatial analysis and applications

Otter Lutra lutra mortality in relation to traffic, and experience with newly established fauna passages at existing road bridges. Lutra

Temporal and spatial influences on road mortality in otters: Conservation implications

Where wildlife and traffic collide: Roadkill rates change through time in a wildlife-tourism hotspot

Wildlife roadkill and COVID-19: A biologically significant, but heterogeneous, reduction

The influence of body-size and diet on road-kill trends in mammals

Raport Roczny "Ogólnopolskiego Rejestru Smiertelnosci Zwierzat na Drogach

Factors influencing mammal roadkills in the agricultural landscape of south-western Poland

Mortality of vertebrates on a road crossing the Biebrza Valley (NE Poland)

Dynamics of wildlife-vehicle collisions on roads of karlovac county

Wildlife-vehicle collisions in Croatia-A hazard for humans and animals

Comparative study of animal carcass survey methods on roads

Spatio-temporal patterns of wildlife-vehicle collisions in a region with a high-density road network

Annual patterns of mammalian mortality on Irish roads

Spatial distribution of road-kills and factors influencing road mortality for mammals in Northern New York State

Transportation Impacts to Wildlife on State Route 37 in Northern San Pablo Bay

Hindering Otter Lutra lutra Road Kills Part 2

Are road-kills representative of wildlife community obtained from atlas data? Hystrix Ital

Behavioural reactions to oncoming vehicles as a crucial aspect of wildlife-vehicle collision risk in three common wildlife species

Relationships between survival and habitat suitability of semi-aquatic mammals

Dynamics in space use of American mink (Neovison vison) in a fishpond area in Northern Germany

The distribution of American mink (Neovison vison) in Greece

Distribution of introduced American mink in the Northern Apennine area (Central Italy). Mammalia 2022

An invasive predator affects habitat use by native prey: American mink and water vole co-existence in riparian habitats

Summer precipitation predicts spatial distributions of semiaquatic mammals

Drying landscape and interannual herbivory-driven habitat degradation control semiaquatic mammal population dynamics

River otter and mink occupancy dynamics in riparian systems

Distribution of invasive muskrats (Ondatra zibethicus) and impact on ecosystem

Decline of invasive muskrat in Lithuania

We thank Deivis Kulvietis and Artūras Kajėnas, both from the Lithuanian Police Traffic Supervision Service, for their help with roadkill data.

The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.