Biodivers Conserv DOI 10.1007/s10531-015-0988-3 ORIGINAL PAPER

Mixed sampling protocols improve the cost-effectiveness of roadkill surveys Aline Saturnino Costa1 • Fernando Ascensa˜o1,2 Alex Bager1

•

Received: 25 March 2015 / Revised: 29 June 2015 / Accepted: 10 August 2015 ! Springer Science+Business Media Dordrecht 2015

Abstract Road mortality due to animal vehicle collisions has serious negative effects on population viability, demanding urgent implementation of mitigation measures where most required. Data from roadkill surveys is generally used to choose the best mitigation measures, but there is a lack of knowledge about the sampling intensity required for surveys to effectively detect roadkill patterns. Improving the cost-effectiveness of road surveys could free up limited resources that, in turn, could be used to obtain a better perception of road impacts and mitigation effectiveness elsewhere. We evaluate the possibility of improving road survey cost-effectiveness by comparing the spatial patterns obtained when using a weekly versus monthly survey and year-round versus seasonal surveys. We analyzed a roadkill dataset, previously collected over 2 years in southern Brazil, and applied two criteria for assessing the effectiveness of two alternative survey protocols: tradeoff between sampling effort and sample size; and similarity in spatial patterns. We found support favoring monthly surveys for those taxa in which roadkill was not markedly temporally clustered, as was the case for larger mammals. However, performing weekly surveys over shorter periods may significantly improve cost-effectiveness

Communicated by P. Ponel. Aline Costa and Fernando Ascensa˜o share the first authorship. Electronic supplementary material The online version of this article (doi:10.1007/s10531-015-0988-3) contains supplementary material, which is available to authorized users. & Aline Saturnino Costa [email protected] & Fernando Ascensa˜o [email protected] 1

Brazilian Center for Road Ecology (CBEE), Ecology Sector, Department of Biology, Federal University of Lavras, Campus Universita´rio, CP 3037, Lavras, MG CEP 37200-000, Brazil

2

Centre for Ecology, Evolution and Environmental Changes (CE3C), Faculdade de Cieˆncias, Universidade de Lisboa, 1749-016 Lisbon, Portugal

123

Biodivers Conserv

for taxa with peaks in mortality if surveys are performed during those periods when roadkill is most likely to occur. We suggest applying a mix of sampling protocols, with intensive surveys during warmer periods for reptiles and other taxa with peak roadkill at these times while, for the rest of the year, a monthly survey could be used to improve detection of rarer species. Keywords Road ecology ! Road monitoring ! Sample size ! Spatial pattern ! Kernel density estimation ! Vertebrates ! Brazil

Introduction Roads have diverse impacts on wildlife at a variety of ecological scales, including changes in animal space use and behavior (Ascensa˜o et al. 2014; Poessel et al. 2014); composition, structure and dynamics of communities and ecosystems (Laurance et al. 2008, 2009); and ultimately may influence the persistence of species at a local or regional extent (Borda-de´ gua et al. 2011; Jaeger et al. 2005). Perhaps the foremost negative effect is roadkill, A which affects virtually every animal species inhabiting the vicinity of a road (Forman et al. 2003). Roadkill can have serious effects on population viability by depleting populations and contributing to genetic structuring (Balkenhol and Waits 2009; Holderegger and Di ´ gua et al. 2014). Hence, Giulio 2010; Jackson and Fahrig 2011; Beebee 2013; Borda-de-A the pervasive nature of roads and their ubiquitous mortality effects demand urgent implementation of mitigation measures where most required. Improving the sustainability of road networks is particularly pressing in regions and countries where these infrastructures will be significantly expanded in the near future to accommodate the increasing human population and concomitant demand for transportation of people, goods and materials (Forman et al. 2003; Laurance et al. 2009, 2014). For example, Brazil has one of the most extensive road networks in the world, comprising over 200,000 km of paved roads (National Agency of Terrestrial Transport—ANTT). Nevertheless, development plans predict the improvement or expansion of 55,000 km of highways and the construction of a further 8000 km, including in currently roadless areas, with inevitable negative consequences for biodiversity (Jaarsma et al. 2006; Selva et al. 2011). Overall, this network expansion demands suitable road planning and mitigation to avoid jeopardizing one of the highest regional biodiversity richness in the world (Brooks et al. 2006; Visconti et al. 2011). Information collected from roadkill surveys is often used to choose the best road mitigation measures (Gunson et al. 2011; Cureton and Deaton 2012; Glista et al. 2009), as well as to assess their effectiveness (Glista et al. 2009; Lesbarreres and Fahrig 2012). The rationale is that road segments or temporal periods with significantly higher roadkill (i.e. hotspots or hot-moments; e.g. Ramp et al. 2005; Hobday and Minstrell 2008; Crawford et al. 2014) should be prioritized for implementation of appropriate measures such as drift fences, road passages, wildlife signage, traffic regulation or even road closure (Lesbarreres and Fahrig 2012; but see Eberhardt et al. 2013). However, intensive road surveys can incur substantial costs and, given limited funds and the need to balance costs between monitoring and mitigation efforts, there is a great incentive to improve the cost-effectiveness of road survey protocols. In fact, funding is a key constraint in environmental impact assessment programs. Thus, once data gathering exceeds that which is sufficient to reliably make a decision regarding road mitigation, resources that could be better spent in other areas such as direct management are wasted.

123

Biodivers Conserv

Cost-effectiveness analyses have been implemented in road ecology studies (Huijser et al. 2009; Rhodes et al. 2014; Polak et al. 2014), but despite the concurrent need for quantitative data to improve road-survey protocols, few studies have sought the optimal sampling approach needed for road surveys of different taxa (Bager and Rosa 2010; Ford et al. 2011; Rosa and Bager 2012). Here, we use a two-year roadkill survey dataset and implicit cost estimates to evaluate the cost-effectiveness of roadkill surveys when using a (1) weekly versus monthly monitoring protocol; and (2) year-round versus periodic monitoring protocols. We considered two types of information deemed as most important in the decision-making process of where to enact mitigation: the sample size and the spatial patterns of roadkill. Collection of a non-representative sample size may bias the perception of the true impact of roadkill on animal populations or preclude analytical procedures aimed at detecting roadkill-related patterns with environmental predictors (e.g., Clevenger et al. 2001; Colino-Rabanal et al. 2010). Likewise, failing to detect the most problematic road stretches may jeopardize mitigation actions by misdirecting resource investment (e.g. passage installation). Hence, we focused the cost-effectiveness evaluation on two criteria: (i) balancing sample size with sampling effort; and (ii) the similarity of roadkill frequency distributions along a road. We provide an analytical framework that integrates both these criteria and that can be applied across many species and landscapes, thus making it of interest to ecologists, conservation biologists and road planners seeking to understand and mitigate the impacts of roads on wildlife populations.

Methods Roadkill dataset We analyzed a roadkill dataset previously collected and described in Rosa and Bager (2012). Road transects were performed in 2002 and 2005 as part of a research program for roadkill mitigation along roads BR392 and BR471, between Pelotas and Santa Vito´ria do Palmar (Fig. 1). The transect covered 117 km in 2002 and, in 2005, the surveyed length was extended 20 km south along the BR 471. These roads have relatively low mean traffic volumes; ca. 10 vehicles per hour (A. Bager, pers. obs.). This region is located in the southernmost Brazilian state of Rio Grande do Sul—an area dominated by palustrine wetlands (Maltchik et al. 2004), with a flattened relief and a high density of lagoons and water bodies (Fig. 1). For this reason, surveys were not strictly regular in time (weekly), as occasionally the roads were closed to traffic due to flooding. Also, days of heavy rain were avoided when carrying out surveys. Thirty-five surveys were performed in 2002 (2.75 ± 1.60 per month), whereas in 2005 roads were surveyed on 42 occasions (3.5 ± 0.67 per month). Survey transects were made by two observers in a car (average speed of ca. 50 km/h), beginning at sunrise and usually lasting until midday. For each roadkill found, observers recorded the date, the GPS position and species (whenever possible). The dataset comprises 2363 records—1313 for 2002 and 1050 for 2005 (Table 1). For further details on the study area and data collection see Bager and Rosa (2011) and Rosa and Bager (2012). Here, we grouped records according to taxon, and species with a high number of records were analyzed separately, including: Water snake Helicops infrataeniatus, Orbigny’s slider turtle Trachemys dorbigni, Argentine black and white tegu Tupinambis merianae,

123

Biodivers Conserv

Fig. 1 Study area location in Brazil and surveyed roads between the city of Pelotas and Taim ecological station (dotted line, total length = 117 km). In 2005, this length was extended by 20 km to cover the area of Taim ecological station. The main land cover classes are shown, derived from Global Land Cover 2010 data (ESA Land Cover CCI; URL: www.esa-cci.org/)

Chestnut-capped blackbird Chrysomus ruficapillus, White eared opossum Didelphis albiventris, Nutria Myocastor coypus, and Capybara Hydrochoerus hydrochaeris. The remaining species were grouped by Class (Table 1). Amphibians were not considered given their low representation (\1 % of total records).

Data analysis First, we analyzed the distribution of roadkill for temporal aggregation of data. Following Malo et al. (2004), we identified peaks (hot-moments) in mortality by comparing the

123

Biodivers Conserv Table 1 Roadkill records per taxa and per year Common name

Scientific name

2002

2005

Total

Water snake

Helicops infrataeniatus

343

186

529

Orbigny’s slider turtle

Trachemys dorbigni

42

68

110

Argentine black and white tegu

Tupinambis merianae

46

45

91

73

129

202

Reptiles (other species) Chestnut-capped blackbird

Chrysomus ruficapillus

Birds (other species) White eared opossum

Didelphis albiventris

Nutria

Myocastor coypus

Capybara

Hydrochoerus hydrochaeris

Mammals (other species) Total

92

37

129

110

78

188

70

123

193

150

68

218 140

86

54

301

262

563

1313

1050

2363

Underlined terms are those used in the text and figures to refer to species. The original dataset comprised 31 species, but here we only used the seven most represented taxa. The remaining species were grouped by Class (reptiles, birds and mammals). Amphibians were not considered given their low representation (\ 1 % of total records)

number of records per survey with the number expected under a random scenario, in which case the likelihood of collisions on each date would show a Poisson distribution. We then measured the effectiveness of alternative road survey protocols according to two criteria. The first criterion focused on the tradeoff between sampling effort and sample size. For any subset of road surveys, we describe this tradeoff as the proportion of roadkill records that were recorded minus the corresponding proportion of surveys that were performed. Positive values indicate a positive tradeoff, i.e. higher cost-effectiveness of the survey protocol, whereas negative values indicate lower cost-effectiveness. For example, if a given subset of surveys included half the number of survey dates (thereby halving the costs) and yet aggregated three-quarters of the records, we have a positive balance in costeffectiveness of 25 %. The second criterion focused on the similarity of spatial patterns between the full dataset and subsets of road surveys. This was measured using kernel density estimation (KDE) of the frequency of kilometric points of roadkill along the road. We derived the kilometric points by measuring the road length from the starting point (near the city of Pelotas) to each GPS location using the plugin LRS—linear reference system (Blazek 2005) in QGIS (QGIS Development Team 2014). The KDE provides a data-driven method for approximating frequency data with probability density functions (Sheather and Jones 1991; Langlois et al. 2012). Moreover, the KDE provides a non-parametric approach to compare pair-wise distributions via a permutation test for shape and location (Bowman and Azzalini 1997), which in our case was represented by sets of kilometric points. The statistical test compares the area between pairs of KDEs (i.e. the full dataset and a subset of data), to that resulting from permutations (n = 999) of the data into random pairs (for further details, see Bowman and Azzalini 1997, 2014). We considered a p value of 0.05 as the threshold to reject the null hypothesis that both distributions were similar. Note that by comparing frequencies we avoided comparing the number of records per road section. A subset of data having a similar distribution to the full dataset has the same location and density of peaks for records along the road. Therefore, any management procedures based

123

Biodivers Conserv

on identification of mortality hotspots will still have the same outcomes for both similar distributions. We compared two alternative roadkill survey protocols against our full dataset, namely monthly and seasonal weekly surveys. For the former, we compared the full dataset (yearround weekly surveys) against hypothetical surveys performed once each month. One random survey date was selected from the pool of dates in each month and the roadkill patterns of the resulting subset were compared to the full dataset. This procedure was repeated for 999 replicates. For the second protocol, the goal was to assess the effectiveness of performing fewer consecutive weekly surveys over a shorter time span than the full year. We compared the survey outputs from the full dataset against subsets of data collected on different starting and finishing dates (using all possible combinations of starting and finishing dates). Costs associated with roadkill surveys are diverse, but mostly relate to fuel. For simplicity, in cost-effectiveness comparisons we assumed a cost of 1 unit of value per survey (i.e. the 2002 and 2005 campaigns had a total cost of 30 and 42 units, respectively). This allowed us to easily quantify the costs of data subsets and relate those with the corresponding effectiveness in terms of the sample size and the similarity in spatial patterns. Comparisons were performed separately for taxa. We also analyzed both years separately to evaluate consistency in our results. Also, most roadkill surveys generally last this timeframe. We only considered data subsets that included at least 25 % of records. Although this is a subjective proportion, based on our experience low sample sizes (associated with low sampling effort) may fail to provide a clear picture of road mortality patterns. KDEs were calculated using the R package ‘sm’ (Bowman and Azzalini 2014). All other analyses and plots were also performed in R (R Development Core Team 2014).

Results Roadkill aggregation analyses allowed us to detect different clustering patterns across the survey periods (Fig. 2). Some species showed marked seasonality in roadkill occurrences, with short-term hot-moments that were particularly evident in warmer periods for water snake and tegu. For water snake, 79 % of records were collected in 14 surveys performed in spring and summer (19 % of sampling effort) whereas, for tegu, 52 % of records were collected in eight surveys (11 % of sampling effort). In addition, surveys performed in April accounted for almost 60 % of all blackbird roadkill records (Fig. 2). It is noteworthy that, for these three species, the hot-moments corresponded across years. In contrast, larger species such as opossum, nutria and capybara showed several mortality peaks throughout the year (particularly in 2005), and these only coincided across both years for the opossum (Fig. 2). Performance of monthly surveys would, theoretically, generally provide 25–50 % of the records of the full dataset in both 2002 and 2005 (Fig. 3, left panel). For water snake and blackbird, the median sample size of 2002 and 2005, respectively, would even be above this interval. Conversely, few subsets of survey dates in 2002 would include 25 % of all blackbird records, which is due to the high temporal clustering of the records for this taxon. Overall, spatial patterns were in agreement (Fig. 3, bottom panel). However, discrepancies were found for water snake, tegu and blackbird for which less than 95 % of subsets had a spatial pattern similar to the weekly dataset, at least in any 1 year.

123

Biodivers Conserv

Fig. 2 Temporal distribution of ‘hot-moment’ roadkill events for the different studied taxa. Only survey dates displaying significant higher mortality are colored. Significance was determined by calculating the probability of recording the number of observations per survey date under a random Poisson distribution scenario. Thresholds were 11, 4, 4, 6, 5, 5, 6, 6, 6 and 12 individuals, respectively, for the water snake to mammals. No hot-moments were detected for nutria and capybara in 2005. Darker grey signifies a higher proportion of data detected (from all observations of the taxa)

Regarding the alternative protocol of seasonal weekly-based surveys, for all taxa we detected subsets of survey dates with a positive tradeoff of reducing the survey effort and the respective proportion of records accumulated, while also providing similar spatial patterns of roadkill densities (Fig. 4, top panel). However, there were different patterns of starting and finishing survey periods for the different taxa, which rarely coincided across both years (Fig. 4, top panel). For example, for water snake, there were only two subsets of dates for 2002, the best of which spanned from mid-February to the end of the year. In contrast, for 2005, the best period spanned from mid-February to mid-October, resulting in

123

Biodivers Conserv Fig. 3 Top:boxplots of the proportion of records represented in monthly subsets of roadkill data. The left and right ‘‘hinges’’ correspond to the first and third quartiles (the 25th and 75th percentiles). ‘Whiskers’ extend from the hinges to the highest and lowest values that are within 1.5 *IQR of the hinge, where IQR is the inter-quartile range. Data beyond the end of the whiskers are outliers and plotted as points. Vertical dotted lines indicates the 25 % threshold. Bottom proportion of subsets of monthly surveys with similar patterns to the full dataset (year-round weekly-based surveys). Longer bars indicate greater similarity in spatial patterns. Vertical dotted line represents the 95 % threshold. In both plots, dark grey represents 2002 and light grey is 2005

76 % of records collected in 64 % of the survey duration. This could represent a 36 % improvement in the cost-effectiveness of the survey protocol for 2005, but does not imply improved efficiency for the 2002 survey. Likewise, in 2002, 39 % of records of tegu were collected in surveys between mid-February and mid-March (13 % of the survey duration), with similar spatial patterns as the full dataset, thus providing an 87 % improvement in cost-effectiveness. However, this short period would fail to detect the roadkill patterns for the 2005 data. Conversely, any survey period covering the mortality hot-moments for blackbirds in April would have retained most of the mortality pattern, as 68 % of records were collected in that survey period. For larger species, there was some correspondence between best survey periods for nutria and capybara, but less so for opossum. Nevertheless, surveys performed between May and October provided 60 % of the records for these three species, encompassing 51 % of the survey effort, thereby representing a potential 9 % improvement in cost-effectiveness. These results suggest that weekly-based surveys can be more cost-effective for a range of species, but there seems to be some inconsistency in terms of the best periods to survey across years.

123

Biodivers Conserv

Fig. 4 Top cost-effectiveness of the weekly roadkill survey protocol: tradeoff between reducing the survey effort and corresponding sample size. Diagonal lines indicate where a reduction in sampling effort represents an equal reduction in the proportion of data collected. Dots above this line have a positive tradeoff. Black dots represent subsets of survey dates that also provide similar patterns of the spatial distribution. Results are presented per year and per taxon. Bottom frequency of starting and finishing dates with positive tradeoff in both criteria of sampling effort and spatial distribution, i.e. of those subsets colored in black (and above the diagonal line) presented in the top panel. Darker density areas are starting dates. For example, the best survey dates for water snake in year 2005 would start around the beginning of February and end by the beginning of November. For some taxa-years, we found no dates fulfilling both criteria

Discussion In this study we assessed two alternative survey protocols aimed at improving the costeffectiveness of roadkill monitoring. We compared the patterns from a two-year dataset of weekly surveys with those obtained by increasing the time-interval between surveys to a monthly-basis and if surveys were performed on a weekly-basis over a shorter time span. We found evidence favoring monthly surveys for those taxa without clear hot-moments in road mortality, whereas performing weekly surveys for shorter time spans (i.e. not all year round) may improve the cost-effectiveness of surveys targeting most taxa, providing the survey period covers these peaks. Widening the interval between surveys may provide an improvement in cost-effectiveness, both in terms of sample size and in the similarity of spatial patterns, particularly for larger mammals. Large mammals typically move long distances to maintain their often large territories, increasing their probability of encountering roads and therefore their mortality risk (e.g. Ascensa˜o et al. 2014). Hence, the greater effectiveness of monthly surveys for this group is probably related to their higher movement rates and greater consistency of roadkill events through time. In fact, although mammals can have distinct periods of higher mortality (Hobday and Minstrell 2008; Grilo et al. 2009; Colino-Rabanal

123

Biodivers Conserv

et al. 2010; Neumann et al. 2012; Poessel et al. 2014), the number of roadkill events was not particularly variable throughout the survey period. These results suggest that for species without a clear seasonal pattern in roadkill events, such as larger mammals, surveys performed on a monthly basis may provide a cost-effective alternative protocol. In our case, carrying out one survey per month (i.e. 12 surveys in any year) could have reduced survey costs by 29 and 37 % of the original costs for 2002 and 2005, respectively. On the other hand, for species with marked mortality peaks such as water snake, tegu or blackbird, monthly-based surveys may fail to effectively detect the true impact of roadkill as spatial patterns may be markedly different. For such species we found evidence of greater effectiveness by weekly surveys to detect roadkill patterns. According to our results, for the smaller species studied here, hot-moments of mortality tend to be concentrated and recurrent across years, while for the larger species peak mortality is less predictable across years and can span longer timeframes. This was likely related to the higher temporal concentration of activity peak events for smaller species. However, the best survey periods were not the same across years (e.g. opossum), or there was no subset of survey time spans with better cost-effectiveness for both years (e.g. blackbird). Hence, significant improvements in cost-effectiveness are possible through weekly surveys over shorter time spans only if they are performed when roadkill is most likely to occur. Considering our results, we suggest that use of mixed protocols throughout the year could provide significant improvements in the cost-effectiveness of road surveys. Intensive surveys should be favored for taxa with strong temporal peaks of activity that generally result in periods of higher mortality. Generally, peaks of activity are associated with climatic conditions, which is particularly true for the herpetofauna or migratory birds. For example, Reading (1998) showed that daily arrival at a breeding pond of sexually mature common toads (Bufo bufo) was highly correlated with mean daily temperatures over the 40 days immediately preceding the main arrival. Similarly, weather patterns at breeding and non-breeding sites can influence the timing of migration for migratory birds such as the American redstart (Setophaga ruticilla) (McKellar et al. 2013) or the barn swallow (Hirundo rustica) (Saino et al. 2004). Likewise, there is a body of literature relating higher rates in roadkill events with specific and predictable climatic conditions. This has been demonstrated for herpetofauna, for which higher mortality rates are expected with the coming of the rainy season (Beaudry et al. 2010; Coelho et al. 2012; Cureton and Deaton 2012; Crawford et al. 2014). Also, roadkill of barn swallows is related to migratory season (Orłowski 2005). Therefore, simple relationships such as these could guide researchers as to when to switch to more intensive survey protocols. Activity peaks are probably more common and easily predicted for small body taxa (e.g. amphibians, reptiles, passerines), and these taxa are also the ones that remain detectable on roads for shorter periods after being hit (Santos et al. 2011). Therefore, we suggest that intensive surveys be scheduled during the periods of higher activity of small-bodied species such as amphibians, reptiles and migratory passerines. During the rest of the year, monthly surveys could be employed to better comprehend mortality patterns and to collect sufficiently large sample sizes (e.g. for larger mammals). Depending on the expected species that are likely to be sampled as roadkill and the climatic patterns in a particular region, a tailored protocol could easily be adapted based on intensive surveys for peak periods of roadkill mortality in smaller species and monthly surveys for the rest of the year. This approach is supported by ad hoc results from 200 data subsets comprising all sampling dates from February through April (i.e. months with higher mortality for reptiles and birds) and one random date for the other 9 months. Nearly all subsets contained over 50 % of the data and described equivalent spatial patterns for the roadkill events of all taxa

123

Biodivers Conserv

(Supplementary Material S1). Overall, this represents a reduction of 50 % of the original survey effort costs of the full dataset (i.e. only 36 of the original 72 surveys would be required). Improvements in roadkill survey protocols are crucial to respond to the urgent demands of countries facing a rapid expansion and improvement of their road networks, as is the case of Brazil. Improving the cost-effectiveness of road surveys facilitates better resource use, including manpower and finances which, in turn, could be used to enhance mitigation of road impacts elsewhere. An important task that is often overlooked in road ecology studies relates to basic information on species in the vicinity of roads such as distribution, abundance and available habitat. This lack of knowledge often precludes inferring the real ´ gua et al. 2014) or behavioral impact of roadkill on population dynamics (Borda-de-A responses toward roads (Grilo et al. 2012). Furthermore, strong statistical inferences are often impossible due a lack of information on ecology before the road was built or improved (BACI design, see Roedenbeck et al. 2007; Van der Grift et al. 2012). Therefore, improved allocation of resources to these tasks may provide decision-makers with better information to support development plans, thereby balancing human needs with conservation biology. One caveat of diminishing the number of surveys relates to species that are rarely recorded or show temporal clusters of mortality lasting only a few hours or days. For example, Santos et al. (2011) described that most animal carcasses persist on roads for only 1 day (note that the focus of their study was on road mortality estimates). Moreover, some species may have low numbers of roadkill events due to their low abundance, ability to avoid incoming vehicles or avoidance behavior toward roads (Jaeger et al. 2005). Species detectability is likely to increase with road sampling effort, and if focal species are rarely observed, lowering the number of surveys may result in low quality data. Therefore, we emphasize that our results are not necessarily applicable to all regions, since they refer to a specific dataset, region, roads and taxa. Further work ought to be devoted to generalizing our results by analyzing different datasets from across the globe and focusing on different species or faunal groups. Acknowledgments We would like to thank Lucas Del Bianco Faria, Clara Grilo, Luis M. Rosalino and ´ gua for their suggestions on the first version of this manuscript. AS and AB were supported Luis Borda-de-A by the project Estrada Viva—RS and financially supported by Fundac¸a˜o O Botica´rio de Protec¸a˜o a` Natureza (O Botica´rio Nature Protection Foundation). The authors have no conflict of interest to declare.

References Ascensa˜o F, Grilo C, LaPoint S, Tracey J, Clevenger AP, Santos-Reis M (2014) Inter-individual variability of Stone Marten behavioral responses to a highway. PLoS ONE 9:e103544 Bager A, da Rosa C (2010) Priority ranking of road sites for mitigating wildlife roadkill. Biota Neotrop 10:2–7 Bager A, da Rosa C (2011) Influence of sampling effort on the estimated richness of roadkilled vertebrate wildlife. Environ Manag 47:851–858 Balkenhol N, Waits LP (2009) Molecular road ecology: exploring the potential of genetics for investigating transportation impacts on wildlife. Mol Ecol 18:4151–4164 Beaudry F, DeMaynadier PG, Hunter ML Jr (2010) Identifying hot moments in road-mortality risk for freshwater turtles. J Wildl Manag 74:152–159 Beebee TJC (2013) Effects of road mortality and mitigation measures on amphibian populations. Conserv Biol 27:657–668 Blazek R (2005) Introducing the linear reference system in GRASS. Int J Geoinf 1:95–100

123

Biodivers Conserv ´ gua L, Navarro L, Gavinhos C, Pereira HM (2011) Spatio-temporal impacts of roads on the Borda-de-A persistence of populations: analytic and numerical approaches. Landsc Ecol 26:253–265 ´ gua L, Grilo C, Pereira HM (2014) Modeling the impact of road mortality on barn owl (Tyto Borda-de-A alba) populations using age-structured models. Ecol Model 276:29–37 Bowman AW, Azzalini A (1997) Applied smoothing techniques for data analysis. Oxford University Press, Oxford Bowman AW, Azzalini A (2014) R package ‘‘sm’’: nonparametric smoothing methods (version 2.2-5.4). URL: http://www.stats.gla.ac.uk/*adrian/sm, http://azzalini.stat.unipd.it/Book_sm. Accessed April 30 2014 Brooks TM, Mittermeier RA, da Fonseca GAB, Gerlach J, Hoffmann M, Lamoreux JF, Mittermeier CG, Pilgrim JD, Rodrigues ASL (2006) Global biodiversity conservation priorities. Science 313:58–61 Clevenger AP, Chruszcz B, Gunson KE (2001) Highway mitigation fencing reduces wildlife-vehicle collisions. Wildl Soc Bull 29(2):646–653 Coelho IP, Teixeira FZ, Colombo P et al (2012) Anuran road-kills neighboring a peri-urban reserve in the Atlantic Forest, Brazil. J Environ Manag 112:17–26 Colino-Rabanal VJ, Lizana M, Peris SJ (2010) Factors influencing wolf Canis lupus roadkills in Northwest Spain. Eur J Wildl Res 57:399–409 Crawford BA, Maerz JC, Nibbelink NP et al (2014) Hot spots and hot moments of diamondback terrapin road-crossing activity. J Appl Ecol 51:367–375 Cureton JCJ, Deaton R (2012) Hot moments and hot spots: identifying factors explaining temporal and spatial variation in turtle road mortality. J Wildl Manag 76:1047–1052 Eberhardt E, Mitchell S, Fahrig L (2013) Road kill hotspots do not effectively indicate mitigation locations when past road kill has depressed populations. J Wildl Manag 77:1353–1359 Ford AT, Clevenger AP, Huijser MP, Dibb A (2011) Planning and prioritization strategies for phased highway mitigation using wildlife-vehicle collision data. Wildl Biol 17:253–265 Forman RTT, Sperling D, Bissonette J et al (2003) Road ecology: science and solutions. Island Press, Washington, DC Glista DJ, DeVault TL, DeWoody JA (2009) A review of mitigation measures for reducing wildlife mortality on roadways. Landsc Urban Plan 91:1–7 Grilo C, Bissonette JA, Santos-Reis M (2009) Spatial–temporal patterns in Mediterranean carnivore road casualties: consequences for mitigation. Biol Conserv 142:301–313 Grilo C, Sousa J, Ascensa˜o F et al (2012) Individual spatial responses towards roads: implications for mortality risk. PLoS ONE 7:e43811 Gunson KE, Mountrakis G, Quackenbush LJ (2011) Spatial wildlife-vehicle collision models: a review of current work and its application to transportation mitigation projects. J Environ Manag 92:1074–1082 Hobday AJ, Minstrell ML (2008) Distribution and abundance of roadkill on Tasmanian highways: human management options. Wildl Res 35:712–726 Holderegger R, Di Giulio M (2010) The genetic effects of roads: a review of empirical evidence. Basic Appl Ecol 11:522–531 Huijser MP, Duffield JW, Clevenger AP et al (2009) Cost–benefit analyses of mitigation measures aimed at reducing collisions with large ungulates in the United States and Canada; a decision support tool. Ecol Soc 14:15 Jaarsma CF, van Langevelde F, Botma H (2006) Flattened fauna and mitigation: traffic victims related to road, traffic, vehicle, and species characteristics. Transp Res Part D: Transp Environ 11:264–276 Jackson ND, Fahrig L (2011) Relative effects of road mortality and decreased connectivity on population genetic diversity. Biol Conserv 144:3143–3148 Jaeger JAG, Bowman J, Brennan J et al (2005) Predicting when animal populations are at risk from roads: an interactive model of road avoidance behavior. Ecol Model 185:329–348 Langlois TJ, Fitzpatrick BR, Fairclough DV et al (2012) Similarities between line fishing and baited stereovideo estimations of length-frequency: novel application of Kernel density estimates. PLoS ONE 7:e45973 Laurance WF, Croes BM, Guissouegou N, Buij R, Dethier M, Alonso A (2008) Impacts of roads, hunting, and habitat alteration on nocturnal mammals in African rainforests. Conserv Biol 22:721–732 Laurance WF, Goosem M, Laurance S (2009) Impacts of roads and linear clearings on tropical forests. Trends Ecol Evol 24:659–669 Laurance WF, Clements GR, Sloan S et al (2014) A global strategy for road building. Nature 513:229–232 Lesbarreres D, Fahrig L (2012) Measures to reduce population fragmentation by roads: what has worked and how do we know? Trends Ecol Evol 27:374–380 Malo JE, Suarez F, Diez A (2004) Can we mitigate animal-vehicle accidents using predictive models? J Appl Ecol 41:701–710

123

Biodivers Conserv Maltchik L, Rolon AS, Guadagnin DL, Stenert C (2004) Wetlands of Rio Grande do Sul, Brazil: a classification with emphasis on plant communities. Acta Limnol Brasiliensia 16:137–151 McKellar AE, Marra PP, Hannon SJ et al (2013) Winter rainfall predicts phenology in widely separated populations of a migrant songbird. Oecologia 172:595–605 Neumann W, Ericsson G, Dettki H, Bunnefeld N, Keuler NS, Helmers DP, Radeloff VC (2012) Difference in spatiotemporal patterns of wildlife road-crossings and wildlife-vehicle collisions. Biol Conserv 145:70–78 Orłowski G (2005) Factors affecting road mortality of the Barn Swallows Hirundo rustica in farmland. Acta Ornithol 40:117–125 Poessel SA, Burdett CL, Boydston EE, Lyren LM, Alonso RS, Fisher RN, Crooks KR (2014) Roads influence movement and home ranges of a fragmentation-sensitive carnivore, the bobcat, in an urban landscape. Biol Conserv 180:224–232 Polak T, Rhodes JR, Jones D, Possingham HP (2014) Optimal planning for mitigating the impacts of roads on wildlife. J Appl Ecol 51:726–734 QGIS Development Team (2014) Quantum GIS Geographic Information System. Open source geospatial foundation. URL: http://qgis.osgeo.org Ramp D, Caldwell J, Edwards KA, Warton D, Croft DB (2005) Modelling of wildlife fatality hotspots along the Snowy Mountain Highway in New South Wales, Australia. Biol Conserv 126:474–490 R Core Team (2014) R: a language and environment for statistical computing. R foundation for statistical computing, Vienna, Austria. URL: http://www.R-project.org/ Reading CJ (1998) The effect of winter temperatures on the timing of breeding activity in the common toad Bufo bufo. Oecologia 117:469–475 Rhodes JR, Lunney D, Callaghan J, McAlpine CA (2014) A few large roads or many small ones? How to accommodate growth in vehicle numbers to minimise impacts on wildlife. PLoS ONE 9(3):e91093 Roedenbeck IA, Fahrig L, Findlay CS, Houlahan JE, Jaeger JAG, Klar N, Kramer-Schadt S, van der Grift EA (2007) The Rauischholzhausen agenda for road ecology. Ecol Soc 12:11 Rosa C, Bager A (2012) Seasonality and habitat types affect roadkill of neotropical birds. J Environ Manag 97:1–5 Saino N, Szep T, Romano M et al (2004) Ecological conditions during winter predict arrival date at the breeding quarters in a trans-Saharan migratory bird. Ecol Lett 7:21–25 Santos SM, Carvalho F, Mira A (2011) How long do the dead survive on the road? Carcass persistence probability and implications for roadkill monitoring surveys. PLoS ONE 6:e25383 Selva N, Kreft S, Kati V, Schluck M, Jonsson BG, Mihok B, Okarma H, Ibisch PL (2011) Roadless and lowtraffic areas as conservation targets in Europe. Environ Manag 48:865–877 Sheather SJ, Jones MC (1991) A reliable data-based bandwidth selection method for kernel density estimation. J Roy Stat Soc Series B (Methodological) 53:683–690 Van der Grift EA, Van der Ree R, Fahrig L et al (2012) Evaluating the effectiveness of road mitigation measures. Biodivers Conserv 22:425–448 Visconti P, Pressey RL, Giorgini D et al (2011) Future hotspots of terrestrial mammal loss. Phil Trans Roy Soc Lond B 366:2693–2702

123