Avian trait-mediated vulnerability to road traffic collisions

Biological Conservation 200 (2016) 122–130

Contents lists available at ScienceDirect

Biological Conservation journal homepage: www.elsevier.com/locate/bioc

Avian trait-mediated vulnerability to road traffic collisions Sara M. Santos a,b,⁎, António Mira a,b, Pedro A. Salgueiro a,b, Pedro Costa a,b, Denis Medinas a,b, Pedro Beja c,d a

CIBIO/InBIO — Research Center in Biodiversity and Genetic Resources, Pólo de Évora, Universidade de Évora, Departamento de Biologia, Mitra, 7002-554 Évora, Portugal UBC — Conservation Biology Unit, Universidade de Évora, Departamento de Biologia, Mitra, 7002-554 Évora, Portugal c EDP Biodiversity Chair, CIBIO/InBIO — Research Center in Biodiversity and Genetic Resources, Universidade do Porto, Campus Agrário de Vairão, Rua Padre Armando Quintas, 4485-661 Vairão, Portugal d CEABN/InBIO, Centro de Ecologia Aplicada “Professor Baeta Neves”, Instituto Superior de Agronomia, Universidade de Lisboa, Tapada da Ajuda, 1349-017 Lisboa, Portugal b

a r t i c l e

i n f o

Article history: Received 11 November 2015 Received in revised form 31 May 2016 Accepted 6 June 2016 Available online xxxx Keywords: Anthropogenic mortality Bird ecology Foraging behaviour Manly's index Road collisions Species traits

a b s t r a c t Collision with vehicles is an important source of bird mortality, but it is uncertain why some species are killed more often than others. Focusing on passerines, we tested whether mortality is associated with bird abundances, and with traits reflecting flight manoeuvrability, habitat, diet, and foraging and social behaviours. We also tested whether the species most vulnerable to road-killing were scarcer near (b500 m) or far (N500–5000 m) from roads. During the breeding seasons of 2009–2011, we surveyed roadkills daily along 50 km of roads, and estimated bird abundances from 74 point counts. After correcting for phylogenetic relatedness, there was strong correlation between roadkill numbers and the abundances of 28 species counted near roads. However, selectivity indices indicated that Blue tit (Parus caeruleus), Blackcap (Sylvia atricapilla) and European goldfinch (Carduelis carduelis) were significantly more road-killed than expected from their abundances, while the inverse was found for seven species. Using phylogenetic generalised estimating equations, we found that selectivity indexes were strongly related to foraging behaviour and habitat type, and weakly so to body size, wing load, diet and social behaviour. The most vulnerable passerines were foliage/bark and swoop foragers, inhabiting woodlands, with small body size and low wing load. The species most vulnerable to road collisions were not scarcer close to roads. Overall, our study suggests that traits provide a basis to identify the passerine species most vulnerable to road collisions, which may be priority targets for future research on the population-level effects of roadkills. © 2016 Elsevier Ltd. All rights reserved.

1. Introduction Roads may have negative impacts on wildlife (Spellerberg, 1998; Fahrig and Rytwinski, 2009; Benítez-López et al., 2010), mainly due to habitat fragmentation and road-related mortality (i.e., roadkills) (Pons, 2000; Erritzoe et al., 2003; Barthelmess and Brooks, 2010; van der Ree et al., 2015). Collisions with vehicles can be particularly detrimental, because hundreds of millions of individuals are killed each year, though the effects of this mortality on the long term persistence of the species most affected are still uncertain (Kociolek et al., 2011; Roger et al., 2011; Borda-de-Água et al., 2014; Loss et al., 2015). Therefore, impact assessment and mitigation should be based on a thorough understanding of the species most vulnerable to road mortality, which is generally very limited (Kociolek et al., 2011).

⁎ Corresponding author at: CIBIO/InBIO — Research Center in Biodiversity and Genetic Resources, Pólo de Évora, Universidade de Évora, Departamento de Biologia, Mitra, 7002-554 Évora, Portugal. E-mail addresses: [email protected] (S.M. Santos), [email protected] (A. Mira), [email protected] (P.A. Salgueiro), [email protected] (P. Costa), [email protected] (D. Medinas), [email protected] (P. Beja).

http://dx.doi.org/10.1016/j.biocon.2016.06.004 0006-3207/© 2016 Elsevier Ltd. All rights reserved.

Birds are often killed due to collisions with vehicles (Erritzoe et al., 2003; Summers et al., 2011; Loss et al., 2015), with roadkills representing up to 5%–10% of overall bird mortality in the Western Paleartic (Møller et al., 2011). Many species are affected, with passerines and owls being among the most commonly reported groups (Erritzoe et al., 2003; Benítez-López et al., 2010; Boves and Belthoff, 2012). However, there is great disparity among studies regarding the number and species killed at different sites, or at the same site over time, though reasons for this are not always clear. One possibility is that the number of road-killed individuals is mainly a consequence of bird abundances near roads, because more individuals of common species are likely to be exposed to collision risk than those of rare species. This idea implies that most collisions should involve species that are naturally abundant in habitats bordering roads (Møller et al., 2011; D'Amico et al., 2015), or species that are attracted by resources provided directly or indirectly by roads such as food, hunting perches, or nesting places (Barrientos and Bolonio, 2009; Morelli et al., 2014; Ascensão et al., 2015). It is possible, however, that road mortality risk is also affected by bird traits, with some species being killed more (or less) frequently than expected from their abundances (Erritzoe et al., 2003; Møller et al., 2011; Cook and Blumstein, 2013). This selective mortality might be important, as it could induce long term changes in bird

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

communities, by progressively depleting the most vulnerable species or groups of species. However, little is known about what traits affect bird susceptibility to road mortality (but see, e.g., Møller et al., 2011; Cook and Blumstein, 2013; Lima et al., 2015), and whether communities actually change due to differential collision risk. Bird morphology is one of the factors that may greatly affect the risk of collision with vehicles. In fact, it is often assumed that birds with heavier bodies, small wings, or a combination of both characteristics (i.e., high wing loadings; Brown and Brown, 2013; DeVault et al., 2015) have higher mortality risk because they are less manoeuvrable (Tennekes, 2009), and have more difficulty in changing direction or reducing flight speed when faced with a moving obstacle (Tennekes, 2009; Martin, 2011; Møller et al., 2011; Brown and Brown, 2013; Legagneux and Ducatez, 2013; DeVault et al., 2015). The type of diet may also be influential, as birds feeding on insects or vertebrates should have higher visual acuity and relative brain size (Garamszegi et al., 2002; Husby and Husby, 2014), and thus presumably lower vulnerability. Also, some birds use foraging techniques that imply flying lower or slower than others, and thus may have higher collision risk (Erritzoe et al., 2003; Kociolek et al., 2011). Other birds make diving manoeuvres or sharp turns, which may also increase collision risk (Erritzoe et al., 2003). Open habitat birds may have lower risk than those of woodland habitats, because they might have higher spatial perception abilities (Lima and Dill, 1990). Finally, flocking behaviour may imply neighbour skills and distance perception (Lima and Dill, 1990; Blumstein, 2006), which could lower their vulnerability (Møller et al., 2011; Cook and Blumstein, 2013). Overall, these studies suggest that functional traits may influence birds' vulnerability to collisions with vehicles, allowing generalization of results and thus yielding practical recommendations to road managers working in different ecological settings (McGill et al., 2006). In this paper we addressed the general idea that roads cause selective bird mortality, and that this selection is due to morphological, behavioural and ecological traits affecting species vulnerability to vehicle collisions. We further investigate whether passerine communities near roads show evidence for depletion of the most vulnerable species. The study was based on a thorough characterization of bird roadkills through daily surveys carried out during the breeding seasons (March to June) of 2009 to 2011, over a 50-km road network, and compared to local bird abundances estimated through point counts. Specifically, we tested the following hypotheses derived from previous studies on bird vulnerability to road collisions (e.g. Erritzoe et al., 2003; Tennekes, 2009; Martin, 2011; Møller et al., 2011; Brown and Brown, 2013; Legagneux and Ducatez, 2013; DeVault et al., 2015): i) there is strong correlation across species between bird abundances and the number of road-killed individuals; ii) there is selective mortality after controlling for the effects of abundance, with higher roadkill risk for species with lower flight manoeuvrability, species that forage on the ground versus woody vegetation, species that feed on plants versus animals, solitary versus flocking species, and forest versus open habitat species; and iii) bird community composition close to roads is depleted in the species most vulnerable to road kills in relation to those far from roads. Results were then used to discuss the effects of roads on bird communities. 2. Materials and methods 2.1. Study area The study was conducted in a 400-km2 area (38° 32′ 24″ to 38° 47′ 33″N; −08° 13′ 33″ to −07° 55′ 45″W) in southern Portugal. The relief is smooth and undulating (150 to 400 m a.s.l.) and the landscape is dominated (N90%) by open woodlands of cork and holm oaks (Quercus suber and Q. rotundifolia) and by farmland (arable land, olive groves, and vineyards). The climate is Mediterranean, with mild winters, and hot and dry summers. Mean temperature varies from 5.8 °C

123

to 12.8 °C during the winter (January), and from 16.3 °C to 30.2 °C during the summer (July). Annual rainfall averages 609.4 mm (Évora 1971– 2000; IPMA, 2014). Roadkill surveys were carried out along four road segments, totalling 50 km: N4 and N114 are national roads (ca. 4000 to 10,000 vehicles/day), and M529 and N370 are narrower roads mostly devoted to regional traffic (b4000 vehicles/day; EP, 2005). All roads are two-lane wide, and central traffic dividers (guardrails) are present in only five road crossings (Fig. 1). 2.2. Roadkill surveys Roadkill surveys were conducted daily from 1st March to 30th June, from 2009 to 2011, by a single observer driving a car at 20–40 km/h and scanning both sides of the road for avian carcasses (details in Santos et al., 2011). The surveys were always carried out by experienced observers (N5 years conducting roadkill surveys), and most (N 80%) were carried out by the same two observers (DM, PC), thereby assuring consistency of procedures and minimising errors due to variation in observer skills. Surveys began at sunrise, corresponding to the daily peak in bird activity (Bibby et al., 2000). A daily schedule was conducted to minimize biases caused by low persistence times of small bird carcasses (Santos et al., 2011). The standard road sampling width corresponded to both lanes and shoulders (paved and unpaved). Every road-killed bird detected was identified in situ or in the laboratory to the lowest possible taxonomic level, and its GPS position was recorded. A total of 2225 bird roadkills of at least 72 species were registered, of which most (93.5%) were Passeriformes (songbird and allies), followed by Strigiformes (owls; 3.7%) and other bird orders (1.7%). Because of this, and because the point counts used to estimate bird abundances around roads (see below) were not adequate to sample owls, aquatic species and birds of prey Bibby et al., 2000), we focused subsequent analyses on Passeriformes. Birds that could not be classified to species level (≈29%) were excluded, except in the case of Crested and Thekla larks, (Galerida cristata and Galerida theklae), which were pooled. In analyses dealing with selective mortality indexes, we considered only roadkill data collected along 23 1-km road sectors at b 500 m from a bird point count (see below). The roadkill species composition at these 23 road sectors was representative of the entire surveyed road (51 sectors), as a multi-response permutation procedure (MRPP; McCune and Grace, 2002) revealed no significant differences in species composition between road sectors with and without bird surveys (A = −0.006, P = 0.871, n = 51). 2.3. Bird abundance data Birds were censused across the study area using 10-minute point counts (Bibby et al., 2000). In each breeding season (April–May) between 2009 and 2011, a total of 74 points was counted twice (early and late breeding season) in the early morning by a single observer (PAS), avoiding windy and rainy days. All birds seen or heard at b250 m from the observer were identified to species level and counted. A total of 31 points was surveyed close to roads (b500 m), and another 43 points were surveyed far from roads (N500 to 5000 m from the road). The points close and far from roads were not paired, because we wanted to characterise bird communities at each road distance class while accounting for the variety of land use types (woodland, farmland, and riparian habitat). Sampling points were placed at least 500 m apart to prevent double counts (Fig. 1). In line with the procedure used for roadkills, only species of Passeriformes were retained in further analyses. Bird abundances at each point were then estimated as the mean number of individuals counted per sampling season, and the abundances close to and far from roads were estimated as the mean number of individuals counted per point within each distance class. Bird species recorded in the study area (names following BirdLife International, 2015) were categorised according to traits potentially

124

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

Fig. 1. Map of the study area in southern Portugal (and Iberian Peninsula), the roads surveyed for bird roadkills and the 74 bird point counts located close (black squares) and far (black stars). Roads are divided into 500-m segments, and they are surrounded by a 500-m buffer (transparent grey strip). The two main land uses are woodland (grey) and farmland (white).

affecting the risk of collision with vehicles (Supplementary material; Table A1). Body weight and wing length were obtained from bibliography (Cramp and Perrins, 1977–1994). Wing loading was estimated by dividing body weight by wing surface area (Blem, 1975; Alerstam et al., 2007). Sociality was defined as solitary/territorial, variable and gregarious. According to the dominant diet, species were classified as granivores, insectivores and omnivores (Cramp and Perrins, 1977-1994; Equipa Atlas, 2008). Foraging guild was defined according to Ehrlich et al. (1994): ground gleaners (pick food from ground surface or plants while walking or clinging); swoopers (snatch prey from the ground in talons after descending from perch); foliage/ bark gleaners (take food items from vegetation or tree trunks, not from the surface of the ground); aerial foragers (capture flying insects while in prolonged continuous flight). Habitat classification was based on the most frequent habitat in which species were recorded in the study area: woodland, farmland, riparian, and generalist (when equally frequent in, at least, two habitats).

2.4. Phylogenetic structure

2.5. Data analysis The relationship across species, between the number of roadkills and mean abundance close to the road, was first investigated through a Pearson correlation coefficient. To account for phylogenetic dependence, we calculated the correlation coefficient using Felsenstein's (1985) independent contrasts method, which is based on an assumption that the traits under study have evolved via Brownian motion along the phylogeny (Paradis, 2012). We then carried out an overall test to evaluate whether at least some species were more (or less) road-killed than expected from their abundances (i.e., the occurrence of selective mortality). This was achieved by comparing the proportional abundances of species in roadkills and in point counts close to roads, using a Pearson chi-square test with P-values obtained through Monte Carlo simulations (10,000 replicates). To assess roadkill selection indices for individual species, we used Manly's standardized selectivity index α (Manly et al., 2003), calculated as: m X α i ¼ ðoi =pi Þ= ðoi =pi Þ i¼1

In analyses involving comparisons across species (see below) we controlled for phylogenetic relatedness, because related species tend to share many traits due to shared evolutionary history (Paradis, 2012). Therefore, they cannot be used in statistical analysis as truly independent observations, and there is the potential for confounding effects resulting from unmeasured conserved traits (Paradis, 2012). To address this problem, we used phylogenetic trees of bird species obtained from “birdTree” http://www.birdTree.org), which provides a tool to extract trees for subsets of a smaller number of species (Jetz et al., 2012). A total of 100 trees were obtained (sample trees from a chosen pseudo-posterior distribution), using data from “Ericson All Species” (Jetz et al., 2012). These trees were then combined in one supertree to be used in subsequent analyses.

with oi = the proportion of road-killed item i, pi = the proportion of item i in the community, m = the number of species in the community. The value of α ranges from 0 to 1, and the value for non-selective roadkill is α = 1/m. Values of α N 1/m indicate that there were more roadkills than expected from abundances near roads (i.e. high mortality risk), whereas values α b 1/m indicate the reverse (Manly et al., 2003). The deviations of α from non-selective road-killing were evaluated by constructing 95% confidence intervals of each mean α, based on 10,000 bootstrap replications of the roadkill and abundance data in each road unit (Efron and Tibshirani, 1993). The deviations of α from nonselective road-killing were only regarded significant when confidence intervals did not include the non-selective value (α = 1/m). Bonferroni

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

confidence intervals were also calculated to adjust for multiple testing over the m different species (Manly et al., 2003). However, because the Bonferonni correction may be too conservative (i.e., it could increase the probability of false negatives, thereby missing significant effects; Nakagawa, 2004), both Bonferroni corrected and uncorrected confidence intervals are presented. The analyses of selectivity excluded all passerines representing less than 1% in both mortality and abundance datasets. To assess which traits increased bird vulnerability to collisions, we regressed the species selectivity indices (dependent variable) with each trait (independent variable). We used phylogenetic generalised estimating equations (GEE; Liang and Zeger, 1986) to account for phylogenetic dependence among observations (Paradis and Claude, 2002; Paradis, 2012). The correlation structure was specified through a correlation matrix that quantifies how much species resemble each other (covariance), and how much they diverged from their common ancestor (variance; Paradis and Claude, 2002). We tested five evolutionary models for defining the correlation structure for each bird trait, including the Brownian motion, and selected the one achieving the best model performance. Because the number of species included in analysis was relatively small (n = 28) and the wing surface area data needed to estimate wing loading was only available for 16 species, we built separate models for each trait. Model fit was verified with plots of residuals versus fitted values (Ballinger, 2004), and Pearson correlations between observed and fitted values. Models were compared with the quasilikelihood information criterion (QIC), which is often used as an alternative to the Akaike's Information Criterion (AIC) in applied longitudinal analysis, and is a representative model selector in GEE methodology (Wang, 2014). The attractive property of QIC is that it allows the selection of the independent variables and the correlation structure simultaneously (Ballinger, 2004; Wang, 2014). The significance of variable coefficients in the model was assessed through t and F tests (Paradis and Claude, 2002). All continuous traits were log transformed to approach normality. Two classes of foraging guild (swoops and foliage/tree gleaning) were clumped to balance the number of species within each class. Differences in the composition of bird communities sampled close and far from roads were estimated using a multi-response permutation procedure (MRPP; McCune and Grace, 2002), followed by an Indicator Species Analysis (Dufrêne and Legendre, 1997) to identify species associated with each area. All analyses and graphical outputs were performed with R 3.0.1 (R Development Core Team, 2013). Phylogenetic analyses were conducted with “ape” and “phangorn” R packages (Paradis et al., 2004; Schliep, 2011). Following recommended statistical good practice (Wasserstein and Lazar, 2016), inferences and interpretation of results considered a combination of P-values, effect sizes, and ecological plausibility.

3. Results 3.1. General patterns During the study period, we recorded 2080 passerine carcasses along the roads surveyed, most of which were European goldfinch (Carduelis carduelis) (9.4%), Blue tit (Parus caeruleus) (8.9%), Corn bunting (Miliaria calandra) (5.7%), House sparrow (Passer domesticus) (5.5%), and Eurasian chaffinch (Fringilla coelebs) (5.5%). Combining breeding bird counts at both road distances, the most abundant species were Corn bunting (12.4% of total passerines counted), Eurasian chaffinch (7.1%), European goldfinch (6.6%), Blue tit (6.0%) and House sparrow (4.4%). There were no significant differences between sampling years in either roadkill (F = 0.011, P = 0.989) or overall species abundances (F = 0.200, P = 0.819).

125

3.2. Selective bird mortality A total of 28 species was recorded at b500 m from point counts and was used in the analysis of selective mortality (Fig. 2). According to our first hypothesis, there was positive correlation across species between abundances close to roads and roadkill numbers (phylogenetically independent contrasts; r = 0.712, P b 0.001). There was also support for the selective mortality hypothesis, with evidence that at least some species were more or less road-killed than expected from their proportions in bird counts (χ2 = 260.11, P b 0.0001). Analysis based on Manly's index indicated significant selectivity in ten species (36%), with more road-kills than expected from abundances recorded for Blue tit, Blackcap (Sylvia atricapilla) and European goldfinch, while fewer road-kills than expected were recorded for Cetti's warbler (Cettia cetti), Spotless starling (Sturnus unicolor), Short-toed tree-creeper (Certhia brachydactyla), Red-rumped swallow (Hirundo daurica), Zitting cisticola (Cisticola jundicis), Winter wren (Troglodytes troglodytes), and Common nightingale (Luscinia megarhynchos) (Fig. 3). More roadkills than expected were also found for Common stonechat (Saxicola torquatus), Great tit (Parus major), Sardinian warbler (Sylvia melanocephala) and European serin (Serinus serinus), but these were only significant without the Bonferroni correction (Fig. 3). Likewise, fewer roadkills than expected were found for Carrion crow (Corvus corone), Eurasian linnet (Carduelis cannabina), Wood nuthatch (Sitta europaea), Crested/ Thekla lark (Galerida spp.), and Eurasian blackbird (Turdus merula), but they lost statistical significance after Bonferroni correction. 3.3. Traits influencing selective mortality The Brownian motion was the evolutionary model achieving the best performance for all traits (lowest QIC), and thus was used for the correlation structure in GEEs. All models had an acceptable fit to observed data, with residuals showing no strong patterns and with the information criteria (QIC) showing small range deviations within comparable models (15.2–17.6) (Table 1). However, these models only supported part of our hypotheses on the likely effects of traits on road collision risk. In contrast to expectations, we found that the selectivity index was not related to wing length, and was inversely related, albeit weakly, to body weight (P = 0.059) and wing loading (P = 0.054), thus suggesting that species with higher flight manoeuvrability were slightly more vulnerable to collision risk (Table 2). Although there was a very strong relation between foraging guild and the selectivity index (Table 1), we found, against expectations, that mortality risk was lowest for ground-feeding species (the reference category in the model) and highest for species with foraging strategies involving swooping or bark/foliage gleaning (Table 2). Evidence for dietary effects was weak (P = 0.073; Table 1), but the tendency was in the expected direction, as mortality risk tended to be highest for granivores (Table 2). Likewise, support for the effect of sociality was weak (P = 0.087), but was also in the expected direction, suggesting higher mortality risk in solitary species. Finally, the strong effect of habitat type (Table 1) was in the expected direction, as the highest risk was found for forest species (Table 2). 3.4. Bird community composition In contrast to our initial hypothesis, there was no evidence for communities sampled near roads to be depleted in the species most vulnerable to road killing. Actually, there were no significant differences in the overall mean abundance of birds in points counted close and far from the road (F = 1.396, P = 0.242), and there were also no significant differences in bird community composition between the two areas (MRPP: A = − 0.006, P = 0.853, n = 51). Still, indicator species analysis revealed an association with areas close

126

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

Fig. 2. Proportion (n = 616 individuals) of the 28 most common bird species recorded road-killed in the sectors used to estimate Manly's indexes for selective road mortality. Surveys were conducted daily from 1st March to 30th June, 2009–2011.

to roads for House sparrow, European goldfinch, Red-rumped swallow, and Black-billed magpie (Pica pica) (species indicator values between 0.52 and 0.61, all P b 0.05), whereas Wood lark (Lullula arborea) and Crested tit (Parus cristatus) were associated with points far from roads (species indicator values between 0.26 and 0.50, all P b 0.05; Supplementary material, Table A2).

4. Discussion 4.1. Abundances and road mortality Our study confirmed that a large number of birds die from collision with vehicles during the breeding season, supporting our hypothesis,

Fig. 3. Manly's selectivity index (α ± 95% bootstrap confidence intervals) for the 28 most road-killed bird species recorded in the roads surveyed. The vertical line indicates the baseline value for non-selectivity (α = 0.036) with values to its right indicating that the species was found road-killed more often than expected from its abundance , and values to its left indicating the reverse. Confidence intervals (error bars) are represented with (grey) and without (black) Bonferroni correction for multiple comparisons.

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

127

Table 1 Summary results of phylogenetic generalised estimating equations (GEE) relating species selective mortality at roads (Manly's selectivity indexes) to morphological, ecological and behavioural traits. QIC: quasilikelihood information criterion; N: number of species used in each model; scale: estimated scale (or dispersion) parameter; pdf: phylogenetic degrees of freedom; F test: F statistic; P: significance of F test; Corr: Pearson correlation). Model

QIC

N

Scale

pdf

F test

P

Corr

Body weight Wing length Wing loading Foraging guild Sociality Habitat Diet

17.587 16.165 –a 15.504 15.587 15.243 16.422

28 28 16 28 28 28 28

0.001 0.001 0.001 0.000 0.001 0.001 0.001

9.0 9.0 6.3 9.0 9.0 9.0 9.0

5.074 2.271 6.985 106.900 3.750 8.012 4.163

0.059 0.176 0.054 0.000 0.088 0.023 0.073

0.204 0.233 0.411 0.726 0.072 0.248 0.209

a

The QIC of the model is not presented because it is not comparable to the others due to smaller sample size.

and observations from previous studies, that the number of individuals of a given species dying on roads is strongly correlated with its relative abundance in nearby (b500 m) areas (Møller et al., 2011; D'Amico et al., 2015). This probably resulted from a simple mass effect, whereby the most abundant species were the most exposed to vehicle collision (D'Amico et al., 2015). It is uncertain, however, whether abundances were shaped mainly by the characteristics of habitats bordering the road, or whether the road itself contributed to affect such abundances by providing resources that would otherwise be absent or scarce (Morelli et al., 2014). Addressing this issue should be the subject of further research, as it may influence the traits of species dying most often on roads (Rytwinski and Fahrig, 2012). 4.2. Trait-mediated vulnerability to road mortality Results also confirmed the hypothesis that some species are killed more (or less) frequently than expected from their abundance, though our expectations based on previous studies (e.g. Møller et al., 2011; Brown and Brown, 2013; Legagneux and Ducatez, 2013) regarding the traits most associated with high mortality risk were often not met. Foraging behaviour was the trait most strongly related to mortality risk, but in contrast to expectations, we found higher risk for foliage/ bark (e.g. Blue tit, Blackcap and European goldfinch) and swoop foragers (e.g., Stonechat), than for ground (e.g., Spotless Starling) and aerial foragers (e.g., Red-rumped Swallow). This may be a consequence of foliage gleaners including species that feed on top of and within shrubs, small trees and large forbs (e.g. thistles often used by European goldfinch), which may be vulnerable to vehicle collision when flying at

low height across open ground from bush to bush. The same may apply to swoop foragers, that capture prey on the ground during short flights from perches, often in woody vegetation, but also when flying from perch to perch and crossing open ground at low height. The vulnerability of swoop foragers may also be related to a common antipredation behaviour, which involves allowing a close approach by the predator and then suddenly escaping (Lima et al., 2015), thereby increasing collision risk by lowering their response time to moving vehicles. Despite the general pattern observed for these groups of species, it is worth noting that mortality was actually lower than expected in some foliage gleaners, including Cetti's warbler and Winter wren. This is probably because they tend to forage within dense woody vegetation (Cramp and Perrins, 1977-1994) and are reluctant to cross large expanses of open ground, which may expose them less to vehicle collisions (Rytwinski and Fahrig, 2012). Likewise, we found lower vulnerability to roadkills for Short-toed tree-creeper, a bark forager that possibly is more reluctant to fly through open areas than other species (Desrocher and Hannon, 2003; Creegan and Osborne, 2005). The relatively lower risk observed for ground foragers was unexpected, as they often walk and fly within the reach of moving vehicles, and may need to exert considerable effort to reach a safe flying height. It is possible, however, that these species have higher surveillance or escape skills due to increased exposure to predators while foraging on the ground, when compared with species relying on tree foliage for cover (Lima, 1990; Watts, 1990; Blumstein, 2006). The low risk observed for aerial foragers such as swallows might also be considered unexpected, because they are often seen flying around and within roads, and typically at low height. These species, however, are very proficient fliers and may

Table 2 Summary of model coefficients of phylogenetic generalised estimating equations (GEE) relating species selective mortality at roads (Manly's selectivity indexes) to morphological, ecological and behavioural traits. Coefficients: estimated coefficients; SE: standard error of estimated coefficients; t test: t statistic; P: significance of t test. Bird traits

Model

Coefficients

SE

t test

P

Body weighta

Intercept body weight Intercept wing length Intercept wing loading Intercept foliage/swoop aerial Intercept variable gregarious Intercept farmland riparian generalist Intercept insectivores omnivorous

0.085 −0.012 0.082 −0.018 0.077 −0.048 0.022 0.044 0.002 0.049 −0.015 −0.018 0.049 −0.017 −0.023 −0.014 0.054 −0.015 −0.026

0.026 0.005 0.033 0.012 0.022 0.018 0.010 0.004 0.012 0.016 0.009 0.010 0.015 0.008 0.009 0.008 0.018 0.014 0.013

3.272 −2.253 2.497 −1.507 3.539 −2.643 2.069 10.323 0.132 3.005 −1.551 −1.921 3.180 −2.265 −2.555 −1.607 3.017 −1.076 −2.022

0.014 0.059 0.041 0.176 0.022 0.054 0.028 0.000 0.899 0.024 0.172 0.103 0.024 0.073 0.051 0.169 0.023 0.323 0.090

Wing lengtha Wing loadinga Foraging guildb

Socialityb

Habitatb

Dietb

a b

Continuous variable. Categorical variable.

128

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

have developed the ability to avoid cars, which could strongly reduce collision risk (Brown and Brown, 2013). The much higher vulnerability found for woodland species in relation to those associated with farmland habitats was in line with our initial expectations. This supports the idea that species living in open habitats may have enhanced spatial perception skills and larger safety distances that increase their awareness of road proximity and avoidance of oncoming cars (Lima and Dill, 1990), which could explain the lower mortality risk of farmland species. The high mortality risk observed for forest species may at least partly reflect the high proportion of foliage gleaners, which were shown to be particularly vulnerable to roadkilling. In contrast to forest species, we found a relatively low mortality risk for species associated with riparian habitats, though they also include a large proportion of foliage gleaners. This may be because some of the most abundant riparian birds in the study area were species such as the Cetti's warbler and the Common nightingale, which live in dense vegetation cover and are reluctant to cross open areas. It could not be ruled out, however, the possibility of the patterns for riparian species being influenced to some extent by the relatively scarce representation of this habitat in the study area. Regarding flight manoeuvrability, we expected in common with other studies, that birds with small body weight, longer wings, and small wing loadings would be less vulnerable to road collisions, due to a higher capacity to escape moving vehicles (Hedenström et al., 2009; Tennekes, 2009; Martin, 2011; Møller et al., 2011; Brown and Brown, 2013; Legagneux and Ducatez, 2013; DeVault et al., 2015). However, we found no effect of wing length, while there was a tendency, albeit weak, for road collision risk actually declining with increasing body size and wing loading. One possible explanation for these unexpected results may be that birds rely on antipredator behaviours to avoid vehicles, but modern vehicles are faster than natural predators (DeVault et al., 2014, 2015). Therefore, the risks of collision might be more related to flight initiation distances and perceptual abilities than to flight manoeuvrability, because even highly proficient fliers may be unable to evade a fast moving vehicle (DeVault et al., 2014, 2015). This would justify the higher risks observed for small bodied species, as they tend to be more tolerant to disturbance and approaching threats (Fernández-Juricic et al., 2001; Blumstein, 2006; Legagneux and Ducatez, 2013), which is often associated with a lower range of spatial perception (Kiltie, 2000) and more difficulty in detecting oncoming vehicles (Garamszegi et al., 2002; Blackwell et al., 2009). In addition, small-bodied species have higher energy demands (Bennett and Harvey, 1987) which may force them to spend more time foraging and travelling between food patches, and thus making dangerous road crossings more often than larger-bodied species. Finally, there may be an interaction between body size and foraging behaviour, as many of the small road-killed birds were foliage gleaners that were shown to be particularly vulnerable to collision with vehicles. These birds are probably killed when moving from one shrub to another across the road, and so they may be in some way distracted and therefore have little reaction to vehicles (Lima et al., 2015). The patterns for wing loading likely resulted, at least to some extent, from its positive correlation with body size (phylogenetically independent contrasts: r = 0.776, P b 0.001). In addition, species with low wing loading and thus higher flight manoeuvrability may have an increased (but false) confidence in their escaping abilities, which may reduce flight initiation distances and thus increase collision risk. Whatever the reasons, our results suggest that high flight manoeuvrability does not decrease bird collision risk with moving vehicles, and may actually indirectly increase this risk. The effects of diet and social behaviour were weak and not statistically significant, but were in the direction expected in our hypotheses. There was a weak tendency for higher risk in species feeding on plants, which may be a consequence of lower visual acuity and relative brain size (Garamszegi et al., 2002; Husby and Husby, 2014). Likewise, there was a small tendency for lower mortality risk in flocking species, which may be related to perceptual abilities and group behaviour

reducing collision risk (Lima and Dill, 1990; Blumstein, 2006; Møller et al., 2011). These effects should be the subject of further research, as they may be confounded by those of other traits with more pronounced effects, such as foraging behaviour and habitat. 4.3. Impacts on bird communities In contrast to our original hypothesis, we found no evidence that bird communities near roads were depleted in the species most vulnerable to road-killing. This is supported by our observations that overall bird abundances and community composition were similar in points sampled close to and far from roads. Also, we found that most individual species were neither associated with roads nor with sites far from roads, irrespective of their high or low vulnerability to vehicle collisions. A few species, however, were more (or less) abundant near roads, but this was probably a consequence of habitat characteristics, rather than roadrelated mortality. For instance, we found that European goldfinch was one of the species most vulnerable to vehicle collisions, but this was also one of the species most strongly associated with points counted close to roads, possibly due to the availability of thistles and other seed-bearing herbs in road verges. The distribution of resources may also explain the higher abundance of other species close to roads (Meunier et al., 1999; Barrientos and Bolonio, 2009; Morelli et al., 2014; Ascensão et al., 2015), including the presence of human structures for nesting (House sparrow, Red-rumped swallow), and the availability of small dead animals due to road collisions (Black-billed magpie). Likewise, Wood lark and Crested tit were the two species most associated with points counted far from roads, but were not particularly vulnerable to road-kills. This distribution was probably also a consequence of differences in habitat availability, as these species are mainly associated with large forested areas in our study area (Godinho and Rabaça, 2010). Overall, these results suggest that local habitats near roads probably affected bird abundances and these in turn affected the number of road-kills, while there was no evidence for road-kills affecting bird abundances, at least for the particular bird communities and ecological setting of our study. 4.4. Limitations and potential shortcomings Our study was based on a number of assumptions and had some limitations, but it is unlikely that they had significant impacts on the main results reported here. A key assumption was that both roadkill surveys and point counts provided a reasonable approximation to the true proportional abundance of different species, thereby allowing meaningful computation of selectivity indices. One problem might be that small birds killed due to collisions with vehicles tend to be underestimated because they decay or are removed quickly, but we believe this problem was largely avoided by conducting daily roadkill surveys in the early morning, thereby guaranteeing the detection of a large proportion of individuals of all sizes killed during the study period (Santos et al., 2011). Another problem could be that some birds are particularly vocal and conspicuous and so may be more easily detected in point counts than unobtrusive species (Alldredge et al., 2011). Although the extent to which this may have affected our results is unknown, it is noteworthy that species with unexpectedly high mortality such as Blue tit, Blackcap and European goldfinch are all very conspicuous and easy to detect in point counts. Thus, it is unlikely that the high selectivity indexes were a consequence of underestimates in their relative abundances. Likewise, it is unlikely that the traits most associated with high vulnerability to roadkills were a consequence of abundance underestimates, as these groups included a wide range of species with varying levels of detectability in point count surveys. Our results apply only to passerines, which have a relatively limited variation in morphological and life-history traits compared to the full range of bird species (Rytwinski and Fahrig, 2012), including mainly species with small sizes, small wings, and low wing loadings. Despite

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

this limitation, our analysis covered species with widely different morphological traits, with body size, for instance, ranging between about 8 g (Zitting cisticola) and 520 g (Carrion crow). Furthermore, passerines accounted for nearly 95% of all bird roadkills recorded in our area, and this is also the group most frequently recorded elsewhere (Erritzoe et al., 2003). Therefore, we believe this limitation does not hinder the validity of our results, though care should be taken when generalizing our results to other landscapes or to non-passerine species that are also frequently road-killed, such as owls (Erritzoe et al., 2003; Guinard et al., 2012; Santos et al., 2013). Another potential problem was that our study focused only on the breeding period, though substantial mortality may also occur at other times of the year. However, this is the season with highest bird roadkill numbers (Garriga et al., 2012), and when mortality may particularly impact on populations by affecting breeding adults (Kuitunen et al., 2003). Nevertheless, it would be important to conduct further studies analysing other periods of the annual cycle, because variations in for instance age-class composition and behaviour (e.g., dispersal movements and winter flocking) could affect species vulnerability to road collisions. In particular, it would be interesting to analyse the post-breeding period, when the presence of large numbers of inexperienced juveniles could greatly affect mortality patterns (Erritzoe et al., 2003; Orlowski, 2008). 4.5. Conservation implications Hundreds of millions of birds die each year due to road collisions (Loss et al., 2015), with our study adding to previous evidence suggesting that this mortality is largely driven by local abundances, and by behavioural, ecological and morphological traits (Møller et al., 2011). However, results from our and other studies suggest that most mortality affects common bird species (D'Amico et al., 2015), including many species that are considered “urban adapters”, which are able to use the resources available in road verges and other highly humanised habitats (McKinney, 2002). In fact, none of the passerine species with high roadkill rates in our study are red-listed in Portugal (Cabral et al., 2007), and species most often killed (such as European goldfinch, Blue tit and Blackcap) are widespread in urban and suburban habitats, though they are also typical of Mediterranean woodlands (Godinho and Rabaça, 2010). No study has yet evaluated whether road mortality may affect the long term persistence of these or other common passerines, and so its conservation relevance is uncertain. However, given the very high number of bird roadkills (Loss et al., 2015), the recent declines of common bird species (Inger et al., 2015), and their importance in ecosystem processes (Dirzo et al., 2014), we suggest that further research is needed on the effects of road mortality on passerine populations. Considering the results of our study on the traits most associated with high mortality risk, we suggest that small-bodied passerines, particularly foliage gleaners and swoop foragers occurring in woodlands crossed by roads, may be priority candidates for such research. Role of the funding source The funding source had no involvement in study design, in the data collection, analysis, and interpretation, in the writing of the manuscript, nor in the decision to submit the paper for publication. Acknowledgements We are grateful to André Lourenço, Clara Ferreira, Paulo Alves, Helena Sabino-Marques, Ana Galantinho, Filipe Carvalho, João Tiago Marques and Carmo Silva for the road surveys and the entire MOVE team. We also thank Carlos Godinho and Pedro Pereira for help in species identification of roadkill carcasses, to Miguel Porto for help in phylogenetic analyses, and Bruno Silva and Márcia Barbosa for help in R script syntax, and to Sasha Vasconcelos for reviewing the English. P.A. Salgueiro and S.M. Santos were funded by grants of the Portuguese

129

Science Foundation (reference SFRH/BD/87177/2012 and SFRH/BPD/ 70124/2010, respectively), and P. Beja was supported by the EDP Biodiversity Chair. We thank the three anonymous reviewers for their constructive comments which helped us to improve the manuscript. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.biocon.2016.06.004. References Alerstam, T., Rosén, M., Backman, J., Ericson, P.G.P., Hellgren, O., 2007. Flight speeds among bird species: allometric and phylogenetic effects. PLoS Biol. 5 (8), e197. http://dx.doi.org/10.1371/journal.pbio.0050197. Alldredge, M.W., Simmons, T.R., Pollock, K.H., 2011. Factors affecting aural detections of songbirds. Ecol. Appl. 17, 948–955. Ascensão, F., LaPoint, S., van der Ree, R., 2015. Roads, traffic and verges: big problems and big opportunities for small mammals. In: van der Ree, R., Smith, D.J., Grilo, C. (Eds.), Handbook of Road Ecology. Wiley-Blackwell, pp. 325–333. Ballinger, G.A., 2004. Using generalized estimating equations for longitudinal data analysis. Organ. Res. Methods 7, 127–150. Barrientos, R., Bolonio, L., 2009. The presence of rabbits adjacent to roads increases polecat road mortality. Biodivers. Conserv. 18, 405–418. Barthelmess, E.L., Brooks, M., 2010. The influence of body-size and diet on road-kill trends in mammals. Biodivers. Conserv. 19, 1611–1629. Benítez-López, A., Alkemade, R., Verweij, P., 2010. The impacts of roads and other infrastructure on mammal and bird populations: a meta-analysis. Biol. Conserv. 143 (1307–131). Bennett, P.M., Harvey, P.M., 1987. Active and resting metabolism in birds: allometry, phylogeny and ecology. J. Zool. 213, 327–363. Bibby, C.J., Burgess, N.D., Hill, D.A., Mustoe, S.H., 2000. Bird Census Techniques. second ed. Academic Press, London. BirdLife International, 2015. The BirdLife checklist of the birds of the world: Version 8. Downloaded from. http://www.birdlife.org/datazone/userfiles/file/Species/ Taxonomy/BirdLife_Checklist_Version_80.zip [.xls zipped 1 MB].[assessed 2nd February 2016]. Blackwell, B.F., Fernández-Juricic, E., Seamans, T.W., Dolan, T., 2009. Avian visual system configuration and behavioral response to object approach. Anim. Behav. 77, 673–684. Blem, C.R., 1975. Geographic variation in wing-loading of the house sparrow. Wilson Bull. 87 (4), 543–549. Blumstein, D.T., 2006. Developing an evolutionary ecology of fear: how life history and natural history traits affect disturbance tolerance in birds. Anim. Behav. 71, 389–399. Borda-de-Água, L., Grilo, C., Pereira, H.M., 2014. Modeling the impact of road mortality on barn owl (Tyto alba) populations using age-structured models. Ecol. Model. 276, 29–37. Boves, T.J., Belthoff, J.R., 2012. Roadway mortality of barn owls in Idaho, USA. J. Wildl. Manag. 76, 1381–1392. Brown, C.R., Brown, M.B., 2013. Where has all the road kill gone? Curr. Biol. 23, R233–R234. Cabral, M.J., Almeida, J., Almeida, P.R., Delliger, T., Fraand de Almeida, N., Oliveira, M.E., Palmeirim, J.M., Queirós, A.I., Rogado, L., Santos-Reis, M. (Eds.), 2007. Livro Vermelho dos Vertebrados de Portugal. Instituto da Conservação da Natureza, Lisboa. Cook, T.C., Blumstein, D., 2013. The omnivore's dilemma: diet explains variation in vulnerability to vehicle collision mortality. Biol. Conserv. 167, 310–315. Cramp, S., Perrins, C.M., 1977-1994. Handbook of the birds of Europe, the Middle East and Africa. The Birds of the Western Palearctic. Oxford University Press, Oxford. Creegan, H.P., Osborne, P.E., 2005. Gap-crossing decisions of woodland songbirds in Scotland: an experimental approach. J. Appl. Ecol. 42, 678–687. D'Amico, M., Román, J., de los Reyes, L., Revilla, E., 2015. Vertebrate road-kill patterns in Mediterranean habitats: who, when and where. Biol. Conserv. 191, 234–242. Desrocher, A., Hannon, S.J., 2003. Gap crossing decisions by forest songbirds during the post-fledging period. Conserv. Biol. 11, 1204–1210. DeVault, T.L., Blackwell, B.F., Seamans, T.W., Lima, S.L., Fernández-Juricic, E., 2014. Effects of vehicle speed on flight initiation by Turkey vultures: implications for bird-vehicle collisions. PLoS One 9 (2), e87944. DeVault, T.L., Blackwell, B.F., Seamans, T.W., Lima, S.L., Fernández-Juridic, E., 2015. Speed kills: ineffective avian escape responses to oncoming vehicles. London B]–>Proc. R. Soc. Lond. B 282, 20142188. http://dx.doi.org/10.1098/rspb.2014.2188. Dirzo, R., Young, H.S., Galetti, M., Ceballos, G., Isaac, N.J., Collen, B., 2014. Defaunation in the anthropocene. Science 345, 401–406. Dufrêne, M., Legendre, P., 1997. Species assemblages and indicator species: the need for a flexible asymmetrical approach. Ecol. Monogr. 67 (3), 345–366. Efron, B., Tibshirani, R.J., 1993. An Introduction to the Bootstrap. Chapman and Hall, Inc., New York (436pp). Ehrlich, P.R., Dobkin, D.S., Wheye, D., Pimm, S.L., 1994. The Birdwatcher's Handbook. Oxford University Press, Oxford. EP, 2005. Recenseamento Do tráfego — Évora. Estradas de Portugal, E.P.E. Equipa Atlas, 2008. Atlas das Aves Nidificantes em Portugal (1999–2005). Instituto Da Conservação Da Natureza e Da Biodiversidade, Sociedade Portuguesa Para O Estudo Das Aves, Parque Natural Da Madeira e Secretaria Regional Do Ambiente e Do Mar. Assírio & Alvim, Lisboa.

130

S.M. Santos et al. / Biological Conservation 200 (2016) 122–130

Erritzoe, J., Mazgajski, T.D., Rejt, L., 2003. Bird casualties on European roads— a review. Acta Ornithologica 38, 77–93. Fahrig, L., Rytwinski, T., 2009. Effects of roads on animal abundance: an empirical review and synthesis. Ecol. Soc. 14, 1–20. Felsenstein, J., 1985. Phylogenies and the comparative method. American Naturalist]– >Am. Nat. 125, 1–15. Fernández-Juricic, E., Jimenez, M.D., Lucas, E., 2001. Alert distance as an alternative measure of bird tolerance to human disturbance: implications for park design. Environ. Conserv. 3, 263–269. Garamszegi, L.Z., Møller, A.P., Erritzøe, J., 2002. Coevolving avian eye size and brain size in relation to prey capture and nocturnality. Proc. R. Soc. Lond. B 269, 961–967. http:// dx.doi.org/10.1098/rspb.2002.1967. Garriga, N., Santos, X., Montori, A., Richter-Boix, A., Franch, M., Llorente, G.A., 2012. Are protected areas truly protected? The impact of road traffic on vertebrate fauna. Biodivers. Conserv. 21, 2761–2774. Godinho, C., Rabaça, J.E., 2010. Birds like it Corky: the influence of habitat features and management of ‘montados’ in breeding bird communities. Agrofor. Syst. 82 (2), 183–195. Guinard, E., Julliard, R., Barbraud, C., 2012. Motorways and bird traffic casualties: carcasses surveys and scavenging bias. Biol. Conserv. 147, 40–51. Hedenström, A., Johansson, L.C., Spedding, G.R., 2009. Bird or bat: comparing airframe design and flight performance. Bioinspir. Biomim. 4, 015001. http://dx.doi.org/10.1088/ 1748-3182/4/1/015001 (13pp). Husby, A., Husby, M., 2014. Interspecific analysis of vehicle avoidance behavior in birds. Behav. Ecol. 25, 504–508. Inger, R., Gregory, R., Duffy, J.P., Stott, I., Voříšek, P., Gaston, K.J., 2015. Common European birds are declining rapidly while less abundant species' numbers are rising. Ecol. Lett. 18 (1), 28–36. IPMA, 2014. Instituto Português do Mar e da Atmosfera, IP, Portugal. Available: http:// www.ipma.pt (accessed 02.07.14). Jetz, W., Thomas, G.H., Joy, J.B., Hartmann, K., Mooers, A.O., 2012. The global diversity of birds in space and time. Nature 491, 444–448. Kiltie, R.A., 2000. Scaling of visual acuity with body size in mammals and birds. Funct. Ecol. 14, 226–234. Kociolek, A.V., Clevenger, A.P., St. Clair, C.C., Proppe, D.S., 2011. Effects of road networks on bird populations. Conserv. Biol. 25, 241–249. Kuitunen, M.T., Viljanen, J., Rossi, E., Stenroos, A., 2003. Impact of busy roads on breeding success in pied flycatchers Ficedula hypoleuca. Environ. Manag. 31, 79–85. Legagneux, P.D., Ducatez, S., 2013. European birds adjust their flight initiation distance to road speed limits. Biol. Lett. 9, 20130417. Liang, K.-Y., Zeger, S.L., 1986. Longitudinal data analysis using generalized linear models. Biometrika 73, 13–22. Lima, S.L., 1990. Protective cover and the use of space: different strategies in finches. Oikos 58, 151–158. Lima, S.L., Blackwell, B.F., DeVault, T.L., Fernández-Juridic, E., 2015. Animal reactions to oncoming vehicles: a conceptual review. Biol. Rev. 90, 60–76. Lima, S.L., Dill, L.M., 1990. Behavioral decisions under the risk of predation: a review and prospectus. Can. J. Zool. 68, 619–640. Loss, S.R., Will, T., Marra, P.P., 2015. Direct mortality of birds from anthropogenic causes. Annu. Rev. Ecol. Evol. Syst. 46, 99–120. Martin, G.R., 2011. Understanding bird collisions with man-made objects: a sensory ecology approach. Ibis 153, 239–254. Manly, B.F.J., McDonald, L.L., Thomas, D.L., McDonald, T.L., Erickson, W.P., 2003. Resource Selection by Animals: Statistical Design and Analysis for Field Studies. second ed. Kluwer Academic Publishers, Dordrecht, The Netherlands. McCune, B., Grace, J.B., 2002. Analysis of Ecological Communities. MjM Software Design, Gleneden Beach, Oregon, USA.

McGill, B.J., Enquist, B.J., Weiher, E., Westoby, M., 2006. Rebuilding community ecology from functional traits. Trends Ecol. Evol. 21, 178–185. McKinney, M.L., 2002. Urbanization, biodiversity, and conservation the impacts of urbanization on native species are poorly studied, but educating a highly urbanized human population about these impacts can greatly improve species conservation in all ecosystems. Bioscience 52 (10), 883–890. Meunier, F.D., Verheyden, C., Jouventin, P., 1999. Bird communities of highway verges: influence of adjacent habitat and roadside management. Acta Oecol. 20 (1), 1–13. Møller, A.P., Erritzoe, H., Erritzoe, J., 2011. A behavioural ecology approach to traffic accidents: interspecific variation in causes of traffic casualties among birds. Zoonoses Res. 32, 115–127. Morelli, F., Beim, M., Jerzak, L., Jones, D., Tryjanowski, P., 2014. Can roads, railways, and related structures have positive effects on birds? — a review. Transp. Res. D 30, 21–31. Nakagawa, S., 2004. A farewell to Bonferroni: the problems of low statistical power and publication bias. Behav. Ecol. 15, 1044–1045. Orlowski, G., 2008. Roadside hedgerows and trees as factors increasing road mortality of birds: implications for management of roadside vegetation in rural landscapes. Landsc. Urban Plan. 86, 153–161. Paradis, E., 2012. Analysis of Phylogenetics and Evolution with R. second ed. Springer, New York (401p). Paradis, E., Claude, J., 2002. Analysis of comparative data using generalized estimating equations. J. Theor. Biol. 218, 175–185. Paradis, E., Claude, J., Strimmer, K., 2004. APE: analyses of phylogenetics and evolution in R language. Bioinformatics 20, 289–290. Pons, P., 2000. Height of the Road Embankment Affects Probability of Traffic Collision by Birds. 47. Bird Study, pp. 122–125. Development Core Team, R., 2013. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (Available from http://www.R-project.org). Roger, E., Laffan, S.W., Ramp, D., 2011. Road impacts a tipping point for wildlife populations in threatened landscapes. Popul. Ecol. 53 (1), 215–227. Rytwinski, T., Fahrig, L., 2012. Do species life history traits explain population responses to roads? A meta-analysis. Biol. Conserv. 147, 87–98. Santos, S.M., Carvalho, F., Mira, A., 2011. How long do the dead survive on the road? Carcass persistence probability and implications for road-kill monitoring surveys. PLoS One 6, e25383. Santos, S.M., Lourenço, R., Mira, A., Beja, P., 2013. Relative effects of road risk, habitat suitability, and connectivity on wildlife roadkills: the case of tawny owls (Strix aluco). PLoS One 8 (11), e79967. http://dx.doi.org/10.1371/journal.pone.0079967 http://dx. doi.org/10.1016/j.envman.2015.07.037. Schliep, K.P., 2011. Phangorn: phylogenetic analysis in R. Bioinformatics 27 (4), 592–593. Spellerberg, I.F., 1998. Effects of roads and traffic: a literature review. Glob. Ecol. Biogeogr. Lett. 7, 317–333. Summers, P.D., Cunnington, G.M., Fahrig, L., 2011. Are the negative effects of roads on breeding birds caused by traffic noise? J. Appl. Ecol. 48, 1527–1534. Tennekes, H., 2009. The Simple Science of Flight, from Insects to Jumbo Jets. Rev. and Expanded Ed. MIT Press, Cambridge. van der Ree, R., Smith, D.J., Grilo, C., 2015. The ecological effects of linear infrastructure and traffic: challenges and opportunities of rapid global growth. In: Smith, D.J., Grilo, C. (Eds.), Van Der Ree, R. Handbook of Road Ecology, Wiley-Blackwell, pp. 1–9. Wasserstein, R.L., Lazar, N.A., 2016. The ASA's statement on P-values: context, process, and purpose. Am. Stat. (just-accepted), 00–00. Wang, M., 2014. Generalized estimating equations in longitudinal data analysis: a review and recent developments. Adv. Stat. 2014, 1–11. Watts, B.D., 1990. Cover use and predator-related mortality in song and savannah sparrows. Auk 107, 775–778.