Anuran road-kills neighboring a peri-urban reserve in the Atlantic Forest, Brazil

Journal of Environmental Management 112 (2012) 17e26

Contents lists available at SciVerse ScienceDirect

Journal of Environmental Management journal homepage: www.elsevier.com/locate/jenvman

Anuran road-kills neighboring a peri-urban reserve in the Atlantic Forest, Brazil Igor Pfeifer Coelho a, b, *, Fernanda Zimmermann Teixeira a, b, Patrick Colombo c, Artur Vicente Pfeifer Coelho d, Andreas Kindel a a NERF e Núcleo de Ecologia de Rodovias e Ferrovias, Departamento de Ecologia, Instituto de Biociências, UFRGS, Av. Bento Gonçalves 9500, CEP 91501-970, CP 15007, Porto Alegre, RS, Brazil b Programa de Pós-graduação em Ecologia, UFRGS, Av. Bento Gonçalves 9500, CEP 91501-970, CP 15007, Porto Alegre, RS, Brazil c Setor de Herpetologia, Laboratório de Sistemática de Vertebrados, Faculdade de Biociências, PUC-RS, Avenida Ipiranga 6681, CEP 90619-900, Porto Alegre, RS, Brazil d Laboratório de Microeletrônica, Instituto de Física, UFRGS, Av. Bento Gonçalves 9500, CEP 91501-970, Porto Alegre, Brazil

a r t i c l e i n f o

a b s t r a c t

Article history: Received 16 March 2012 Received in revised form 28 June 2012 Accepted 5 July 2012 Available online

Mortality from road-kills may figure among the important causes of decline in amphibian populations and species extinctions worldwide. Evaluation of the magnitude, composition, and temporal and spatial distributions of amphibian road-kills is a key step for mitigation planning, especially in peri-urban reserves. Once a month for 16 months, we surveyed, on foot, a 4.4 km section of state road ERS-389 bordering the Itapeva reserve in the southern Atlantic Forest. We recorded 1433 anuran road-kills and estimated a mortality rate of 9002 road-kills/km/year. The species most often recorded were the largest ones: Leptodactylus latrans, Rhinella icterica, Leptodactylus gracilis and Hypsiboas faber; 54.5% of the carcasses could not be identified. Anuran mortality was concentrated in summer, and was associated with temperature, rainfall and photoperiod. Leptodactylus road-kills were strongly influenced by vehicle traffic, probably because of its high abundance during the entire study period. Road-kill hotspots differed for anurans as a group and for single species, and we found an association among spatial patterns of mortality and types of land cover, distance from the nearest waterbody, roadside ditches, and artificial light. Traffic should be banned temporarily during periods of high mortality, which can be forecasted based on meteorological data. A comprehensive mitigation approach should take into account hotspots of all anuran records, and also of target species for selecting locations for amphibian passages and fencing. Roadside ditches, artificial waterbodies, and conventional street lights should be reduced as much as possible, since they may represent ecological traps for anuran populations. Ó 2012 Elsevier Ltd. All rights reserved.

Keywords: Amphibian Atlantic Forest Itapeva Road-kill hotspots Mitigation planning

1. Introduction In recent decades, population declines and species extinctions in amphibian assemblages have been documented in many regions of the world, and amphibians were recognized as one of the most threatened animal groups in the ongoing biodiversity crisis (Stuart et al., 2004). Nowadays, 41% of amphibian species that have been adequately evaluated are considered to be threatened, a much higher proportion than estimates for birds and mammals (IUCN, 2011). Singly or in combination, climate change, habitat loss, habitat fragmentation and splitting, invasive species introduction, increases in UV-B radiation and infectious diseases (especially chytridiomycosis), chemical pollution, and overexploitation are the main factors threatening amphibians (Semlitsch, 2003; Becker * Corresponding author. NERF e Núcleo de Ecologia de Rodovias e Ferrovias, Departamento de Ecologia, Instituto de Biociências, UFRGS, Av. Bento Gonçalves 9500, CEP 91501-970, CP 15007, Porto Alegre, RS, Brazil. Tel.: þ55 51 32423874. E-mail address: [email protected] (I.P. Coelho). 0301-4797/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jenvman.2012.07.004

et al., 2007; Pounds et al., 2006; Young et al., 2001). Wildlifevehicle collisions, one of the main direct causes of animal mortality worldwide (Forman and Alexander, 1998), have also been associated with declines of amphibian populations (Fahrig et al., 1995) and may represent the greatest threat for some species (Elzanowski et al., 2009; Hels and Buchwald, 2001). The majority of studies concerning amphibian road-kills have been conducted in temperate zones and developed countries, and knowledge about estimated mortality, spatio-temporal patterns, and associated factors is limited to these regions (Puky, 2006). The high mortality rates observed (e.g. Ehmann and Cogger, 1985, estimated that 4.45 million anurans are killed on roads each year in Australia) may also be occurring in other regions. South America is the continent with the highest number and density of amphibian species (Duellman, 1999), and harbors 35.9% of threatened amphibian species (IUCN, 2011). However, local evaluations of amphibian road mortality are still rare in South America, although this source of mortality may have an important impact on populations. Cairo and Zalba (2007), for example, demonstrated that

18

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

a road through the Argentinean Pampas has a significant impact on the local population of Melanophryniscus sp., with an estimated annual road-kill mortality of up to 5.9% of the population. Effectively managed systems of protected areas are thought to be crucial for protecting amphibians from human activities (Young et al., 2001). However, the ongoing extension of human occupation over landscapes leads to expansion of infrastructure such as roads near protected areas, which may compromise the functionality and effectiveness of these areas for conservation. Since roads near urban areas carry heavy and continuous vehicle traffic, the impact of neighboring roads on small and peri-urban reserves may be extremely severe (Ramp et al., 2006). It is critical to evaluate spatial and temporal patterns of road-kill in these situations, to provide information to use in planning mitigation measures, especially for threatened groups such as amphibians. We describe the species composition and estimate the magnitude of anuran road-kill on a road bordering a peri-urban reserve in the Atlantic Forest of southern Brazil. We analyze the spatial and temporal distributions of anuran road-kills in general, and for the individual species that were most often recorded (Leptodactylus latrans, Rhinella icterica, Leptodactylus gracilis and Hypsiboas faber), identifying which factors are responsible for the higher rates of accidents at certain times and locations. We discuss the impact of mortality in a regional context, and describe mitigation measures that could be implemented locally, based on our findings. 2. Material and methods 2.1. Study area We conducted this study near the Itapeva State Park (PEVA), in the southern part of the Atlantic Forest Biosphere Reserve, Brazil (Fig. 1). The local climate is subtropical humid, with an annual mean temperature of 19  C and annual mean precipitation of 1400 mm. The reserve covers an area of 1000 ha, encompassing the last preserved coastal ecosystems in the region, including mobile dunes, “restinga” forest (medium sized trees and shrubs adapted to sandy soils), marshy forest, marshes, peatlands, and also humanmade marshy and dry grasslands and transitional formations (Lindeman et al., 1975; Kindel, 2002). This high ecosystem diversity in a small area has led to a relatively high anuran species richness (28 spp.) for this latitude (Colombo et al., 2008). State road ERS-389 bordering the PEVA is 11 m wide, with one lane in each direction and no shoulder. Buses and trucks are prohibited, and the speed limit is 80 km/h. Traffic volume is high in summer and low in winter, with an annual mean of 2132 vehicles per day, maximum of 4431 in January, and minimum of 1252 in July (2002 data; DAER RS www.daer.rs.gov.br). 2.2. Road-kill data We evaluated anuran road-kills monthly, between July 2002 and October 2003. In the morning, two observers on foot surveyed (one in each road lane) a 4410 m section of ERS-389, neighboring the PEVA. Anuran carcasses found were identified, their locations determined with a GPS receiver, and removed from the road. 2.3. Variables associated with road-kill 2.3.1. Temporal variables We tested the relationship between temporal variations of anuran road-kills and: TEMPERATURE, mean of daily mean temperature; RAINFALL, total precipitation; PHOTOPERIOD, mean number of hours of sunlight per day; VEHICLE TRAFFIC, total number of vehicles; and VEHICLE SPEED, percentages of vehicles above 40, 60 and 80 km/h.

Fig. 1. Road ERS-389 (black line) bordering the peri-urban Itapeva Reserve (PEVA, white line), Torres municipality, southern Brazil.

Each variable was calculated for the entire periods of 20, 10, 5 and 2 days before the day of the survey in each month, in order to evaluate the scale dependence of these variables. We obtained data for temperature and precipitation from the meteorological station in Torres municipality (Meteorology National Institute, INMET www. inmet.gov.br), 3.5 km distant from the study area. Daily sunlight hours were obtained from the Brazilian National Observatory (http:// euler.on.br/ephemeris/index.php). To calculate traffic volume and vehicle speed, we used data from a speed monitor located on the road section surveyed (DAER RS - www.daer.rs.gov.br). 2.3.2. Spatial variables We investigated the association between road-kills and the areas of the different classes of land cover: WATER, rivers and lakes; MARSH; URBAN, houses and other buildings; AGRICULTURE; MARSHY FOREST; GRASSLAND; PASTURE; SHRUBS; PAVED ROAD; UNPAVED ROAD; MANAGED WOODS, woods with alien and/or native fruit trees without an understory; EXPOSED SOIL; and SILVICULTURE, Eucalyptus sp. Other spatial variables evaluated were: ROAD DITCHES, length of ditches up to 20 m from the road; ARTIFICIAL LIGHT, light intensity considering the number of street lights and their distance from the road; VEHICLE SPEED, mean vehicle speed on the road segment; WATERBODY DISTANCE, distance to the nearest perennial waterbody. We divided the road into 63 segments of 70 m each, and calculated the spatial variables for each of the 64 points that delimited these segments. We used ArcMap 9.2 software (ESRI, Redlands, California) and a Quick Bird image (pixel 0.6  0.6 m) classified by visual interpretation (Dobrovolski, 2006) to obtain the classes of land cover in buffers of 50, 100 and 200 m radius centered on each of the 64 points. The variables ROAD DITCHES, ARTIFICIAL LIGHT, and VEHICLE SPEED were measured in the field with a GPS receiver and calculated for each point, using a buffer of 50 m radius

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

on GPS TrackMaker software (www.trackmaker.com). Mean vehicle speed on each road segment was estimated using the mean speed of vehicles (n ¼ 19) that we followed by car, recording the speed during the route with a GPS.

2.4. Statistical analyses 2.4.1. Temporal distribution of road-kills We evaluated the temporal distribution of road-kills for all anurans and for species with high numbers of records, through circular analysis (Zar, 1999). Using ORIANA 2.02 software (Kovach, 2004), we converted the months (n ¼ 16) into angles (22.5 grade intervals) and used the number of road-kills in each month as the frequency of each angle, obtaining a mean angle that represents the average period with the highest mortality during the time interval evaluated. The Rayleigh Uniformity test was performed to evaluate the significance of the average period in relation to a uniform distribution of records (Kovach, 2004; Zar, 1999). We also estimated the intensity of road-kill concentration by the length (r) of the average period, which varies from 0 (uniform dispersion) to 1 (concentration of records in the same direction). 2.4.2. Spatial distribution of road-kills The Ripley’s K statistic is used to evaluate the dispersion of events on different spatial scales (Clevenger et al., 2003; Levine, 2000; Ripley, 1981). We used a modified Ripley’s K (Coelho et al., 2008) in SIRIEMA v1.0 software (www.ufrgs.br/biociencias/ siriema) to determine the scales on which road-kills were significantly aggregated in space. To define the different scales evaluated, we used an initial radius of 50 m and increments of 50 m for each step. To evaluate the significance of possible aggregations, we subtracted the observed K values from the mean obtained in 1000 simulations of random road-kill distributions for each scale. Values above the confidence limits (99%) obtained from the simulations indicate scales with significant aggregations (Levine, 2000). We located the road stretches with high mortality using 2D HotSpot Identification analysis in SIRIEMA v1.0 software. In this analysis, the road is divided into segments of the same length (we used 63 segments of 70 m each, delimited by 64 points). A circle with radius r (we used a 50 m radius) is centered on the first of these 64 points, and all road-kill events inside the circle area are summed. This sum is multiplied by a correction factor that considers the length of the road inside the circle in this position. Then, the circle is centered on the next point and the sum is again computed and multiplied by the correction factor. This procedure is repeated for all 64 points delimiting the road segments, resulting in an aggregation intensity value for each road segment:

Hi ðrÞ ¼ 2r=Ci ðrÞ

n X

fij

j¼1

where: Hi(r) ¼ aggregation value for point i considering scale r; n ¼ number of road-kill events; r ¼ defined radius; i ¼ point on road; j ¼ road-kill event; Ci(r) ¼ road length inside the circle, with r radius centered on point i; fij ¼ index equal to 0 if j is outside the circle with r radius centered on i, or equal to 1 if j is inside this area. To evaluate the significance of the aggregation intensity, for each point the following function is used:

IðrÞ ¼ Hi ðrÞ  HsðrÞ; where: Hs(r) ¼ mean of H values in 1000 simulations of random distribution of the events. Values for aggregation intensity above the upper confidence limit (90%) indicate significant road-kill hotspots.

19

2.4.3. Association between road-kills and explanatory variables We transformed temporal and spatial explanatory variables into X0 ¼ log (X þ 1). We included in the multiple regression analyses only those variables that showed a significant linear correlation (Pearson’s correlation coefficient, P < 0.05) with road-kills. We used two multiple regression approaches (path analysis and model selection using AIC) to evaluate the relationships between explanatory variables and all anuran road-kills, and separately, between explanatory variables and road-kills of the species recorded most often (L. latrans, R. icterica, L. gracilis and H. faber). We carried out path analyses to identify the relative importance of the direct and indirect influences of each temporal variable on the temporal distribution of mortality, since some explanatory variables have a known influence on other explanatory variables (PHOTOPERIOD and RAINFALL on TEMPERATURE). First, we tested in which period (20, 10, 5 or 2 days) each variable showed the highest significant association with road-kills, using the highest Pearson’s correlation coefficient obtained. We calculated path coefficients, their significance, and the model fit using the maximum-likelihood estimate in AMOS 5.0.1 software (Arbuckle, 2007). Model fit was tested by Chi-square test. The normality of each variable was evaluated observing skewness and kurtosis values, and multivariate normality was tested using kurtosis (Mardia’s coefficient values below 1.96 indicate that there is no significant kurtosis; Mardia, 1970). For H. faber, we also show the standardized regression coefficients in a path diagram, although TEMPERATURE did not show a significant correlation with roadkills, and was not used in the model. Since there are no known relationships among the spatial explanatory variables (and we had no hypotheses about these relationships to test), we performed a model selection using the second-order Akaike information criterion (AICc) to evaluate which variables were more related to road-kills in space. Considering 64 sample units (the 64 points that limit the road segments of 70 m each), we first evaluated on which scales (buffers with a radius of 50, 100 and 200 m) the land cover classes showed the highest Pearson’s correlation coefficient with the intensity of road-kill aggregation. Correlations between selected spatial variables were not superior to 0.75. All possible models for each response variable and its set of explanatory variables were ranked according to lower AICc scores, and the Akaike weight for each model (um) was estimated using SAM software (Rangel et al., 2010). Models with an AICc difference lower than two when compared with the lowest one were considered as plausible models (Burnham and Anderson, 2002). The importance of each explanatory variable was calculated by the sum of the Akaike weights over the subset of models were P the variable was included ( ui), and the standardized regression coefficients (bi) were estimated using model averaging across all models (Burnham and Anderson, 2002). 2.4.4. Estimate of annual mortality magnitude We performed two estimates of anuran annual mortality using multiple regression, considering the predictive variables identified in the path analysis as having a significant direct influence on anuran road-kills. Because carcasses decompose with time and vehicle traffic, and are removed by scavengers, they remain on the road for a very short time, less than 24 h (Teixeira, 2011), or on the order of 1 or 2 (Santos et al., 2011) or 3.2 days (Gerow et al., 2010). For this reason, we defined pseudo-samplings for each day (total of 365 days) or every three days (total of 121 days) during the period from July 2002 through June 2003. We estimated the road-kill number for each pseudo-sampling using the values from predictive variables for those days in the multiple regression equation obtained in the path analysis. The sum of all road-kills observed during the real sampling days and the road-kills estimated for

20

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

pseudo-sampling days was used to estimate the total annual mortality. For the purpose of comparing between methods, we also estimated mortality using the equation proposed by Teixeira (2011):

N ¼ nplTR where N ¼ total number of road-kills observed, n ¼ number of surveys, p ¼ detection probability, l ¼ mortality rate (road-kill number per time unit), and TR ¼ characteristic time for carcass removal. This equation was developed in order to correct for bias in road-kill estimates caused by carcass removal and searcher detection. We used TR value (0.964 days) obtained for anurans by Teixeira (2011) on a road near ERS-389 (approximately 30 km to the southwest) and a detection probability of 0.9 (close to the maximum detectability value, considering that the survey was performed on foot). We calculated the daily mortality rate (l) and multiplied it by 365 to generate an annual estimate. 3. Results We recorded a total of 1433 anuran road-kills, belonging to 13 species and six families (Table 1). The majority of records (54.7% of the total) were not identified to species, and 35% could not be identified even to family level, due to carcass disintegration. The family Leptodactylidae had the highest number of records (35.7%). The species most often recorded were L. latrans (18.9%), R. icterica (6.8%), L. gracilis (6.7%) and H. faber (6.6%). 3.1. Temporal distribution of road-kills and related factors Anuran road-kills were significantly concentrated (Rayleigh’s Uniformity Test, Z ¼ 601.9, P < 0.001) in the summer months, with a mean concentration period estimated in January (r ¼ 0.64; Fig. 2a). Individual species mortality was also concentrated in some periods (Fig. 2b, c, d, e). For L. latrans (Z ¼ 203.4, P < 0.001; r ¼ 0.86), Table 1 Species of road-killed anurans on state road ERS-389 bordering the Itapeva Reserve, southern Brazil. Taxon Bufonidae Rhinella icterica Bufonidae NI Microhylidae Elachistocleis bicolor Hylidae Hypsiboas faber Hypsiboas guentheri Scinax fuscovarius Scinax granulatus Scinax squalirostris Sphaenorhynchus aff. surdus Hylidae NI Leptodactylidae Leptodactylus gracilis Leptodactylus latrans Leptodactylidae NI Cycloramphidae Odontophrynus maisuma Leiuperidae Physalaemus gracilis Physalaemus lisei Leiuperidae NI

N (%) 98 (6.8) 14 (0.9) 12 (0.8) 95 21 5 1 2 1 60

(6.6) (1.4) (0.3) (0.007) (0.14) (0.07) (4.1)

97 (6.7) 272 (18.9) 152 (10.6) 27 (1.8) 15 (1.0) 3 (0.2) 46 (3.2)

Anura NI

512 (35.7)

Total Species richness

1433 13

NI e not identified.

L. gracilis (Z ¼ 38.5, P < 0.001; r ¼ 0.63) and H. faber (Z ¼ 71.9, P < 0.001; r ¼ 0.87) there were also more records during summer, and the mean periods with the highest road-kill occurrence estimated were in February (Fig. 2b), between December and January (Fig. 2d) and in December (Fig. 2e), respectively. R. icterica mortality was more evenly distributed through the year (Z ¼ 35.8, P < 0.001; r ¼ 0.60), with the highest mortality in October 2002 (Fig. 2c). Path diagrams that describe the relationships among temporal variables and road-kills are presented in Fig. 3. Models for Anura (c2 ¼ 0.11; DF ¼ 1; P ¼ 0.73), L. gracilis (c2 ¼ 2.25; DF ¼ 2; P ¼ 0.32) and H. faber (c2 ¼ 0.99; DF ¼ 2; P ¼ 0.60) showed a good fit to the data. The model for L. latrans showed a significant Chi-square, which indicates a lack of satisfactory model fit (c2 ¼ 7.23; DF ¼ 1; P ¼ 0.007). However, we used this model for interpretation because the variables were highly correlated, making a poor fit and leading to a Type II error of inference in the Chi-square test. For the R. icterica model, there was no value of positive degrees of freedom to calculate the model fit. Temperature showed the strongest effect on anuran road-kills (standardized path coefficient ¼ 0.63), followed by precipitation (0.45). Vehicle traffic did not show a significant direct effect on anuran road-kills in this evaluation, considering the other variables. Photoperiod, although it did not show a significant direct effect on mortality, showed an indirect effect through the strong influence on temperature (0.88). Although vehicle traffic did not show a significant influence on anurans as a whole, evaluations of individual species showed that it influenced road-kills of L. gracilis and L. latrans (standardized path coefficient ¼ 0.56 and 0.50, respectively). Temperature directly influenced only L. latrans road-kills, but photoperiod (which showed an indirect relationship with temperature in all cases, except for R. icterica since temperature did not enter into the model) directly influenced R. icterica (0.38) and H. faber (0.57). Precipitation, although it showed a significant direct relationship with anurans in general, for the individual species separately, significantly influenced only the mortality of R. icterica (0.51). 3.2. Spatial distribution of road-kills and related factors Road-kill spatial distribution for anurans in general and for the species separately on state road ERS-389 was non-random for almost all scales evaluated (Fig. S1 in Supplementary material), indicating that the records were concentrated in some locations along the road section. The location of road-kill hotspots for anurans and for the individual species differed along the road (Fig. 4), indicating that different factors may be more important for mortality occurrence in each case. The model selection provided 18 models for anuran road-kills (R2 z 0.53), five for L. latrans (R2 z 0.64), three for R. icterica (R2 z 0.47), 13 for L. gracilis (R2 z 0.56), and three for H. faber (R2 z 0.23), which could be considered as similarly plausible according to their AICc (Table S1). Table 2 shows the selected variables for predicting spatial distributions of road-kills, its relative importance and the standardized regression coefficients. Anuran mortality was mainly associated with WATERBODY DISTANCE P P ( ui ¼ 0.966, b ¼ 0.48) and WATER ( ui ¼ 0.921, b ¼ 0.28). ROAD P P DITCHES ( ui ¼ 0.999, b ¼ 0.45), PASTURE ( ui ¼ 0.969, b ¼ 0.37), P and UNPAVED ROAD ( ui ¼ 0.991, b ¼ 0.40) were the most important variables related to L. latrans road-kills. ARTIFICIAL LIGHT P P ( ui ¼ 0.987, b ¼ 0.44), WATERBODY DISTANCE ( ui ¼ 0.738, P b ¼ 0.28), MANAGED WOODS ( ui ¼ 0.703, b ¼ 0.23), SILVICULP P TURE ( ui ¼ 0.674, b ¼ 0.25), and GRASSLAND ( ui ¼ 0.625, b ¼ 0.20) were identified as the main predictors for R. icterica mortality. L. gracilis road-kills were mostly related to SHRUBS P P ( ui ¼ 0.999, b ¼ 0.40), ROAD DITCHES ( ui ¼ 0.966, b ¼ 0.33), and

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

21

Fig. 2. Temporal distribution of road-kills on ERS-389: (a) anurans, (b) Leptodactylus latrans, (c) Rhinella icterica, (d) Leptodactylus gracilis and (e) Hypsiboas faber. The mean angle (solid black line) and standard error (dotted lines) of the circular analysis are presented.

P WATER ( ui ¼ 0.962, b ¼ 0.31). The models for H. faber, which explained a small amount of variation in road-kills, showed AGRIP CULTURE ( ui ¼ 0.956, b ¼ 0.34) and MANAGED WOODS P ( ui ¼ 0.765, b ¼ 0.27) as the main explanatory variables.

3.3. Anuran road-kill estimates The predictor variables TEMPERATURE and RAINFALL were used in the equation X ¼ 6.69 þ 1.17*Rainfall þ 4.67*Temperature

22

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

Fig. 3. Path diagrams of variables influencing the temporal distribution of road-kills: (a) anurans, (b) Leptodactylus latrans, (c) Rhinella icterica, (d) Leptodactylus gracilis and (e) Hypsiboas faber. Wide arrows indicate significant path coefficients (P < 0.05), and dashed arrows indicate non-significant paths. Width of arrows is proportional to the effect of the relationship (standardized path coefficients, numbers shown). Numbers in parentheses represent the proportion of variance of each endogenous variable explained by the model (R2). UN ¼ unexplained variance (error terms). Non-causal covariations were considered in the analyses but are not represented graphically.

(R2 ¼ 0.82, F2,13 ¼ 36.16, P < 0.001) to estimate anuran annual mortality on roads. We estimated that the annual road-kill was on the order of 36,011 (9002 road-kills/km/year) when considering one day of pseudo-sampling interval, and 11,233 individuals (2547 road-kills/km/year) when considering three days. Based on the equation proposed by Teixeira (2011), the estimated mortality was 37,683 anurans per year (9420 road-kills/km/year). 4. Discussion 4.1. Anuran road-kill magnitude and composition Amphibians are the vertebrate group with the highest mortality from vehicle collisions in many regions (Attademo et al., 2011; Carvalho and Mira, 2010; Gerow et al., 2010; Glista et al., 2008; Gryz and Krauze, 2008; Turci and Bernarde, 2009), and may be the group most impacted by this factor worldwide. Observed annual road-kill rates per kilometer for anurans may be on the order of 3.4 (gravel

and dirt roads, Coleman et al., 2008), 245 (this study), or 246 (Attademo et al., 2011), reaching 21,940 road-kills/km/year (Goldingay and Taylor, 2006), although comparisons among these studies are biased due to differences in sampling intervals. Observed road-kill values, however, are certainly underestimates of real mortality. Mortality estimates based on observed road-kills, especially considering observers’ detection capabilities and carcass persistence (Gerow et al., 2010; Santos et al., 2011; Teixeira, 2011), are needed to evaluate the magnitude of this impact and to compare the impacts among different roads and regions. The method that we used to estimate road-kill, using predictive variables, yielded a similar result to that obtained using observers’ detection and carcass persistence values (9002 and 9420 road-kills/ km/year, respectively), indicating that it can be used in the absence of experiments on detection and removal rates, when predictive variables that explain a substantial amount of temporal variability in mortality are identified (in this case, 82%). The only study that also estimated anuran mortality rate is Gerow et al. (2010), who

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

23

Fig. 4. Road-kill intensity of aggregation (black line) and 90% confidence limits (light-gray lines) along 4.41 km of ERS-389: (a) anurans, (b) Leptodactylus latrans, (c) Rhinella icterica, (d) Leptodactylus gracilis and (e) Hypsiboas faber. Values above the upper confidence limit indicate significant hotspots of mortality.

estimated 159.3 road-kills/km/year in the desert of Saguaro National Park, Arizona. A large part of the anuran assemblage in the PEVA region is subject to road-kill impact; 46% of the known species in the reserve and 56% of the species that occur in habitats near the road have been recorded as road-kill (Colombo et al., 2008). The anuran species most vulnerable to road-kills are those with terrestrial

habits, high vagility, long migration distances during the breeding season, low rates of movement, philopatry to reproductive sites, and a daily movement pattern that coincides with peaks of traffic flow (Carr and Fahrig, 2001; Hels and Buchwald, 2001; Orlowski et al., 2008; Sillero, 2008). Species with high numbers of records on ERS-389 are common, and are generalists capable of exploiting a wide range of environmental conditions. Silva et al. (2007) found

24

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

Table 2 P Relative importance based on Akaike weights ( ui) and standardized regression coefficients (bi) estimated from model averaging for predictor spatial variables of road-kill intensity of anurans as a group, and of single species. Those variables without values in each column were not included in the analysis, due to the lack of significant linear correlation with road-kills, or were not selected in the plausible models subset (see Methods). Anura P

ui

ARTIFICIAL LIGHT ROAD DITCHES VEHICLE SPEED WATERBODY DISTANCE WATER (100 m) WATER (200 m) URBAN (100 m) URBAN (200 m) PASTURE (50 m) PASTURE (200 m) GRASSLAND (50 m) GRASSLAND (200 m) PAVED ROAD (100 m) UNPAVED ROAD (100 m) UNPAVED ROAD (200 m) MANAGED WOODS (50 m) MANAGED WOODS (200 m) SHRUBS (50 m) SILVICULTURE (200 m) MARSHY FOREST (200 m) AGRICULTURE (100 m) AGRICULTURE (200 m) EXPOSED SOIL (200 m)

bi (SE)

L. latrans P

ui

0.999 0.29 0.966 0.921

0.10 (33.8) 0.48 (8.7) 0.28 (3.7)

0.259

0.06 (0.47)

0.536

bi (SE)

R. icterica P

ui

0.44 (8.4)

0.326 0.738

0.15 (7.2) 0.28 (0.83)

0.45 (20.3) 0.20 (0.7)

0.237 0.387

0.157 (14)

0.969

0.435

0.479

0.16 (3)

0.33 (14.7) 0.25 (8.2) 0.18 (0.5)

0.962

0.31 (2.6)

0.716

0.26 (2.6)

0.687 0.825

ui

bi (SE)

0.266

0.06 (0.34)

0.765

0.27 (0.83)

0.001 (3)

0.40 (3.1) 0.03 (0.11) 0.703

0.05 (0.5)

0.966 0.563 0.373

0.212 0.991 0.229

bi (SE)

0.20 (0.8)

0.10 (42.2) 0.19 (1.5)

0.251

ui

H. faber P

0.04 (0.3)

0.37 (6) 0.625

0.319 0.657

bi (SE)

0.987

L. gracilis P

0.14 (0.45)

0.19 (0.68) 0.24 (1)

similar results to ours, in a study in another area in the southern Atlantic Forest, where L. latrans was also the species with the highest number of road-kills. In other studies of anurans in Brazil, the genus Rhinella showed the highest road-kill frequency (reviewed by Dornas et al., 2012); it is also the genus most often recorded on roads in Santa Fé province, Argentina (Attademo et al., 2011). The predominance of Leptodactylus and Rhinella in road-kill records may be due to the large body size of species in these genera, which increases the probability of detection. The majority of carcasses that we could not identify to species level (54.7% of total) were small anurans. The high proportion of non-identified carcasses may mask the species that are most often road-killed, and may even mask the possible mortality of rare and threatened species. Small sized species (such as Sphaenorhynchus aff. surdus, Pseudopaludicola falcipes, and Physalaemus sp.) can be poorly recorded due to rapid carcass removal. Also, forest-interior species (such as Leptodactylus cf. araucaria, Haddadus binotatus, and Ischnocnema henselii) will barely be road-killed in the study area. We recommend that amphibian road-kill surveys be done early in the morning (or even twice a day), considering that most road-kills probably occur at night; and also on consecutive days, in order to improve detection and species identification. Unfortunately, since there are no estimates of species abundances in the area, the demographic effects of this mortality on anuran populations could not be estimated. Although most of the species recorded are still common, the long-term effects of roadkills need to be evaluated and mitigation measures should be implemented, since common species may rapidly become rare, especially those that are specialized on widespread environmental conditions (Lindenmayer et al., 2011).

0.674 0.262 0.238 0.267

0.23 (0.69) 0.25 (0.6) 0.07 (0.6) 0.01 (0.23) 0.06 (0.31)

0.999

0.40 (0.23)

0.671

0.20 (0.83)

0.586

0.21 (1.2)

0.568 0.325

0.22 (0.39) 0.10 (0.2)

0.956

0.34 (0.82)

dermal permeability and ectothermy are the main physiological characteristics driving anuran behavior and ecology, making these organisms’ activity pattern, movements, and locomotor performance closely related to temperature and water availability (Wells, 2007). Photoperiod is also important in the regulation of anuran activity rhythms (Wells, 2007), and is recognized as a determinant factor in anuran reproductive activity in southern Brazil (Both et al., 2008). Other studies demonstrated a positive relationship of anuran roadkill to temperature and precipitation (Attademo et al., 2011), with road-kill peaks during adult migration periods in spring, and juveniles during summer and autumn (Gryz and Krauze, 2008; Orlowski, 2007; Orlowski et al., 2008). Leptodactylus road-kills were strongly influenced by vehicle traffic. In the study area, Leptodactylus spp. were abundant throughout the year, even outside the breeding season (P. Colombo, pers. obs.), probably making variations in vehicle traffic the main determinant for the number of road-kills. Mazerolle (2004) observed that even small temporal changes in the traffic volume on a road (e.g., 5e26 vehicles/h) may increase mortality for some amphibian species (e.g., Bufo americanus); whereas for others (e.g., Pseudacris crucifer), the number of road-kills decreases with increasing traffic, probably because the animals avoid the road when traffic is heavy. Vehicle traffic may influence amphibian road-kills in different ways, increasing the number of accidents until a traffic volume threshold is reached (see the relationship between moose Alces alces and traffic described by Seiler, 2005) in cases when species do not avoid the road, or decreasing the species’ occurrence when anurans avoid roads with moderate to heavy traffic. The only study that directly evaluated road avoidance for an anuran species is that of Bouchard et al. (2009), who concluded that high mortality of Rana pipens was due to failure to avoid roads or traffic.

4.2. Factors related to temporal distribution of road-kills 4.3. Factors related to spatial distribution of road-kills Variables associated with organisms’ activity (especially temperature, but also precipitation and photoperiod) were the most important factors related to the temporal distribution of anuran roadkills overall, and to road-kills of L. latrans, R. icterica and H. faber. High

Spatial variations of vertebrate road-kills may be a result of the local abundance of species (Seiler, 2005), local traffic volume (Orlowski and Nowak, 2006), and/or factors that increase the

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

probability of collision, such as vehicle speed (Hobday and Minstrell, 2008; Seiler, 2005). We did not evaluate the spatial influence of vehicle traffic because we considered that traffic volume did not vary along the road, since it is a short section connected to a few dirt roads used by local residents. However, we detected an association of anuran and single-species road-kills with types of land cover, waterbody distance, road ditches, and artificial light, indicating the importance of local, fine-scale anuran abundance for the occurrence of mortality hotspots. Local anuran abundance is related to landscape composition and structure (Gray et al., 2004; Knutson et al., 1999), among other factors. For members of the family Ranidae, forest area and wetland area are related to road-kill increases, while the presence of residences and urban area are related to decreases in road-kill numbers (Glista et al., 2008). Langen et al. (2009) showed that the presence and configuration of wetlands within 100 m from roads are associated with anuran road-kill hotspots, and are more important than traffic volume. Mortality of the common toad (Bufo bufo) is associated with local population abundance and with the area of waterbodies adjacent to roads, but not with traffic volume (Orlowski, 2007). However, Orlowski (2007) suggested that the lack of a relationship is due to a long-term cumulative effect of mortality on roads with heavy traffic, causing the local decline of populations, as in the case of Pelobates fuscus in Sweden (Nyström et al., 2007). The association between R. icterica road-kills and artificial light may be a result of high concentrations of individuals of this species searching for prey near street lights. Some anuran species use areas with artificial lights to forage, and common toads (B. bufo) concentrate close to street lights to capture insects (Buchanan, 2006). We found no important spatial relationship between roadkills and vehicle speed, probably due to high average speed and small local variations (74 km/h and 63e83 km/h, respectively). Although vehicle speed may not be important for anuran perception of the vehicle and escape (as it is for other animals of larger size and/or with a high capacity for movement), it may be an important factor for mortality caused by sudden changes of pressure. Even without being smashed by vehicle wheels, anurans can die by blowout, a high-pressure wave created by vehicles traveling above 30 km/h (Hummel, 2001). Therefore, one question that needs to be evaluated is the relative roles of blowout and collision in causing mortality. If the proportion of anuran road-kills due to blowout is high compared to mortality due to collision, road sections with a mean vehicle speed lower than 30 km/h could show less mortality, and speed reduction may also be an efficient mitigation measure for amphibians. 5. Conclusions and management implications Anuran mortality was concentrated in summer, and was associated with temperature, rainfall and photoperiod. Implementation of temporary mitigation measures should be a priority during hot moments, that is, periods with high concentration of road-kills (Beaudry et al., 2010). In addition to the periods already identified when anuran road-kill rates are high along ERS-389, short periods when conditions favor anuran activity (related to temperature, precipitation and photoperiod) also must be considered and forecasted using predictive models based on meteorological data, using an adaptive road-mitigation approach. The identification of daily hot moments is needed to implement mitigation approaches such as temporary traffic bans and mobile speed monitors (if speed reduction is effective; see 4.3. “blowout” discussion). Road-kill hotspots differs for anurans as a group and for single species, and the spatial pattern of mortality is associated with factors that can be managed to diminish this impact, such as types of land cover, distance from the nearest waterbody, roadside

25

ditches, and artificial light. A comprehensive mitigation approach should take into account mortality hotspots for anurans as a group and also for target species in selecting where to implement measures, such as passages with guide fencing and barrier fences designed especially for amphibians (Woltz et al., 2008). Spatial models considering landscape characteristics associated with mortality hotspots may also be used to determine the location of mitigation measures in other ERS-389 sections and nearby roads. The formation of road ditches and construction of artificial waterbodies near roads must be avoided since they may represent ecological traps (Robertson and Hutto, 2006) for anuran populations. Fewer street lights along roads and/or the use of light bulbs that are less attractive to insects must also be considered in order to reduce road-kill of particular anuran species. Acknowledgments We thank all colleagues who assisted in data collection, especially Graciela B. Horn and Sofia Zank. Ricardo Dobrowolski provided a classified satellite image made available from the Secretaria Estadual do Meio Ambiente do Rio Grande do Sul (SEMARS). SEMA provided lodging in the PEVA during field work. This study was partially supported by the Instituto de Biociências of UFRGS. Appendix A. Supplementary material Supplementary material associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jenvman. 2012.07.004. References Arbuckle, J.L., 2007. Amos 16.0 User’s Guide. CopyrightÓ 1995e2007 by Amos Development Corporation. Spring House, PA, USA. Attademo, A.M., Peltzer, P.M., Lajmanovich, R.C., Elberg, G., Junges, C., Sanchez, L.C., Bassó, A., 2011. Wildlife vertebrate mortality in roads from Santa Fe Province, Argentina. Rev. Mex. Biodivers. 82, 915e925. Beaudry, F., Demaynadier, P.G., Hunter Jr., M.L., 2010. Identifying hot moments in road-mortality risk for freshwater turtles. J. Wildl. Manage. 74 (1), 152e159. Becker, C.G., Fonseca, C.R., Haddad, C.F.B., Batista, R.F., Prado, P.I., 2007. Habitat split and the global decline of amphibians. Science 318, 1775e1777. Both, C., Kaefer, I.L., Santos, T.G., Cechinm, S.T.Z., 2008. An austral anuran assemblage in the Neotropics: seasonal occurrence correlated with photoperiod. J. Nat. Hist. 42 (3e4), 205e222. Bouchard, J., Ford, A.T., Eigenbrod, F.E., Fahrig, L., 2009. Behavioral responses of northern leopard frogs (Rana pipiens) to roads and traffic: implications for population persistence. Ecol. Soc. 14 (2), 23. Buchanan, B.W., 2006. Observed and potential effects of artificial night lighting on anuran amphibians. In: Rich, C., Longcore, T. (Eds.), Ecological Consequences of Artificial Night Lighting. Island Press, Washington, DC, pp. 192e220. Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multimodel Inference. Springer, New York. Cairo, S.L., Zalba, S.M., 2007. Effects of a paved road on mortality and mobility of red bellied toads (Melanophryniscus sp.) in Argentinean grasslands. AmphibiaReptilia 28, 377e385. Carr, L.W., Fahrig, L., 2001. Effect of road traffic on two amphibian species of differing vagility. Conserv. Biol. 15 (4), 1071e1078. Carvalho, F., Mira, A., 2010. Comparing annual vertebrate road kills over two time periods, 9 years apart: a case study in Mediterranean farmland. Eur. J. Wildl. Res. 57 (1), 157e174. Clevenger, A.P., Chruszcz, B., Gunson, K.E., 2003. Spatial patterns and factors influencing small vertebrate fauna road-kill aggregations. Biol. Conserv. 109, 15e26. Coelho, I.P., Kindel, A., Coelho, A.V.P., 2008. Road-kills of vertebrate species on two highways through the Atlantic Forest Biosphere Reserve, southern Brazil. Eur. J. Wildl. Res. 54, 689e699. Coleman, J.L., Ford, N.B., Herriman, K., 2008. A road survey of amphibians and reptiles in a bottomland hardwood forest. Southeast. Nat. 7 (2), 339e348. Colombo, P., Kindel, A., Vinciprova, G., Krause, L., 2008. Composição e ameaças à conservação dos anfíbios anuros do Parque Estadual de Itapeva, município de Torres, Rio Grande do Sul, Brasil. Biota Neotrop. 8 (3), 229e240. Dobrovolski, R., 2006. A paisagem do Parque Estadual de Itapeva, RS, e seu entorno: padrões, processos e fatores direcionadores. Dissertation, Federal University of Rio Grande do Sul, Porto Alegre, Brazil.

26

I.P. Coelho et al. / Journal of Environmental Management 112 (2012) 17e26

Dornas, R.A.P., Kindel, A., Bager, A., Freitas, S.R., 2012. Avaliação da mortalidade de vertebrados em rodovias no Brasil. In: Bager, A. (Ed.), Ecologia de Estradas: Tendências e Pesquisas. Editora UFLA, Lavras. Duellman, W.E., 1999. Patterns of Distribution of Amphibians: a Global Perspective. The Johns Hopkins University Press, Baltimore. Ehmann, H., Cogger, H., 1985. Australia’s endangered herpetofauna: a review of criteria and polices. In: Grigg, G., Shine, R., Ehman, H. (Eds.), Biology of Australian Frogs and Reptiles. Surrey Beatty & Sons and Royal Zoological Society of New South Wales, Sydney.  ska, J., Urban, R., 2009. Elzanowski, A., Ciesiolkiewicz, J., Kaczor, M., Radwan Amphibian road mortality in Europe: a meta-analysis with new data from Poland. Eur. J. Wildl. Res. 55, 33e43. Fahrig, L., Pedlar, J.H., Pope, S.E., Taylor, P.D., Wegner, J.F., 1995. Effect of road traffic on amphibian density. Biol. Conserv. 73, 177e182. Forman, R.T.T., Alexander, L.E., 1998. Roads and their major ecological effects. Annu. Rev. Ecol. Evol. Syst. 29, 207e231. Gerow, K., Kline, N.C., Swann, D.E., Pokorny, M., 2010. Estimating annual vertebrate mortality on roads at Saguaro National Park, Arizona. HumaneWildlife Interact. 4 (2), 283e292. Glista, D.J., DeVault, T.L., DeWoody, J.A., 2008. Vertebrate road mortality predominantly impacts amphibians. Herpetol. Conserv. Biol. 3 (1), 77e87. Goldingay, R.L., Taylor, B.D., 2006. How many frogs are killed on a road in North-east New South Wales? Aust. Zool. 33 (3), 332e336. Gray, M.J., Smith, L.M., Leyva, R.I., 2004. Influence of agricultural landscape structure on a Southern High Plains, USA, amphibian assemblage. Landscape Ecol. 19, 719e729. Gryz, J., Krauze, D., 2008. Mortality of vertebrates on a road crossing the Biebrza Valley (NE Poland). Eur. J. Wildl. Res. 54, 709e714. Hels, T., Buchwald, E., 2001. The effect of road kills on amphibian populations. Biol. Conserv. 99, 331e340. Hobday, A.J., Minstrell, M.L., 2008. Distribution and abundance of roadkill on Tasmanian highways: human management options. Wildl. Res. 35, 712e726. Hummel, D., 2001. Amphibienschutz durch Geschwindigkeitsbegrenzung e eine aerodynamische Studie. Nat. Landsch. 76 (12), 530e533. IUCN, 2011. IUCN Red List of Threatened Species. Version 2011.1. http://www. iucnredlist.org. Downloaded on 30 October 2011. Kindel, A., 2002. Diversidade e estratégias de dispersão de plantas vasculares da floresta paludosa do faxinal, Torres, RS. Thesis, Federal University of Rio Grande do Sul, Porto Alegre, Brazil. Knutson, M.G., Sauer, J.R., Olsen, D.A., Mossman, M.J., Hemesath, L.M., Lannoo, M.J., 1999. Effects of landscape composition and wetland fragmentation on frog and toad abundance and species richness in Iowa and Wisconsin. U.S.A. Conserv. Biol. 13 (6), 1437e1446. Kovach, W.L., 2004. Oriana for Windows. Version 2.02. Kovach Computer Services, Pentraeth, Wales, UK. Langen, T.A., Ogden, K.M., Schwarting, L.L., 2009. Predicting hot spots of herpetofauna road mortality along highway networks. J. Wildl. Manage. 73 (1), 104e114. Levine, N., 2000. CrimeStat: a Spatial Statistics Program for the Analysis of Crime Incident Locations. Ned Levine & Associates, Annandale, Virginia. Lindeman, J.C., Irgang, B., Porto, M.L.P., Loscheiter, M.L., Batista, L.R.M., 1975. Estudos botânicos no Parque Estadual de Torres, Rio Grande do Sul, Brasil. Levantamento florístico da planície do curtume, da área de Itapeva e da área colonizada. Iheringia Bot. 21, 15e52. Lindenmayer, D.B., Wood, J.T., McBurney, L., MacGregor, C., Youngentob, K., Banks, S.C., 2011. How to make a common species rare: a case against conservation complacency. Biol. Conserv. 144, 1663e1672. Mardia, K.V., 1970. Measures of multivariate skewness and kurtosis with applications. Biometrika 57, 519e530. Mazerolle, M.J., 2004. Amphibian road mortality in response to nightly variations in traffic intensity. Herpetologica 60 (1), 45e53.

Nyström, P., Hansson, J., Mansson, J., Sundstedt, M., Reslow, C., Broström, A., 2007. A documented amphibian decline over 40 years: possible causes and implications for species recovery. Biol. Conserv. 138, 399e411. Orlowski, G., Nowak, L., 2006. Factors influencing mammal roadkills in the agricultural landscape of south-western Poland. Pol. J. Ecol. 54 (2), 283e294. Orlowski, G., Ciesiolkiewicz, J., Kaczor, M., Radwanska, J., Zywicka, A., 2008. Species composition and habitat correlates of amphibian roadkills in different landscapes of south-western Poland. Pol. J. Ecol. 56 (4), 659e671. Orlowski, G., 2007. Spatial distribution and seasonal pattern in road mortality of the common toad Bufo bufo in an agricultural landscape of south-western Poland. Amphibia-Reptilia 28, 25e31. Pounds, J.A., Bustamante, M.R., Coloma, L.A., Consuegra, J.A., Fogden, M.P.L., Foster, P.N., La Marca, E., Masters, K.L., Merino-Viteri, A., Puschendorf, R., Ron, S.R., Sánchez-Azofeifa, G.A., Still, C.J., Young, B.E., 2006. Widespread amphibian extinctions from epidemic disease driven by global warming. Nature 439, 161e167. Puky, M., 2006. Amphibian road kills: a global perspective. In: Irwin, C.L., Garrett, P., McDermott, K.P. (Eds.), Proceedings of the 2005 International Conference on Ecology and Transportation. Center for Transportation and the Environment, North Carolina State University, Raleigh, NC, pp. 325e338. Ramp, D., Wilson, V.K., Croft, D.B., 2006. Assessing the impacts of roads in periurban reserves: road-based fatalities and road usage by wildlife in the Royal National Park, New South Wales, Australia. Biol. Conserv. 129, 348e359. Rangel, T.F., Diniz-Filho, J.A.F., Bini, L.M., 2010. SAM: a comprehensive application for spatial analysis in macroecology. Ecography 33, 46e50. Ripley, B.D., 1981. Spatial Statistics. John Wiley and Sons, New York. Robertson, B.A., Hutto, R.L., 2006. A framework for understanding ecological traps and an evaluation of existing evidence. Ecology 87 (5), 1075e1085. 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 (9), e25383. http://dx.doi.org/10.1371/ journal.pone.0025383. Seiler, A., 2005. Predicting locations of mooseevehicle collisions in Sweden. J. Appl. Ecol. 42, 371e372. Semlitsch, R.D., 2003. Amphibian Conservation. Smithsonian Institution, Washington, D.C. Sillero, N., 2008. Amphibian mortality levels on Spanish country roads: descriptive and spatial analysis. Amphibia-Reptilia 29, 337e347. Silva, M.O., Oliveira, I.S., Cardoso, M.W., Graf, V., 2007. Road kills impact over the herpetofauna of Atlantic Forest (PR-340, Antonina, Paraná). Acta Biol. Par. 36 (1e2), 103e112. Stuart, S.N., Chanson, J.S., Cox, N.A., Young, B.E., Rodrigues, A.S.L., Fischman, D.L., Waller, R.W., 2004. Status and trends of amphibian declines and extinctions worldwide. Science 306, 1783e1786. Teixeira, F.Z., 2011. Fauna atropelada: estimativas de mortalidade e identificação de zonas de agregação. Dissertation, Federal University of Rio Grande do Sul, Porto Alegre, Brazil. Turci, L.C.B., Bernarde, P.S., 2009. Vertebrados atropelados na Rodovia Estadual 383 em Rondônia, Brasil. Biotemas 22 (1), 121e127. Wells, K.D., 2007. The Ecology and Behavior of Amphibians. The University of Chicago Press, Chicago. Woltz, H.W., Gibbs, J.P., Peter, K., Ducey, P.K., 2008. Road crossing structures for amphibians and reptiles: informing design through behavioral analysis. Biol. Conserv. 141, 2745e2750. Young, B.E., Lips, K.R., Reaser, J.K., Ibáñez, R., Salas, A.W., Cedeño, J.R., Coloma, L.A., Ron, S., La Marca, E., Meyer, J.R., Muñoz, A., Bolaños, F., Chaves, G., Romo, D., 2001. Population declines and priorities for amphibian conservation in Latin America. Conserv. Biol. 15 (5), 1213e1223. Zar, J.H., 1999. Biostatistical Analysis. Prentice-Hall, New Jersey.