1 Amphibia-Reptilia 33 (2012): Interactive effects of roads and weather on juvenile amphibian movements Mireille Gravel 1, Marc J. Mazerolle 2,, Marc-André Villard 1 Abstract. We investigated whether paved roads adjacent to 16 ponds acted as barriers to movements of juvenile wood frogs (Lithobates sylvaticus), green frogs (Lithobates clamitans), mole salamanders (Ambystoma laterale, A. maculatum), and American toads (Anaxyrus americanus) in eastern New Brunswick, Canada. Using pitfall traps and drift fences, we recorded captures of juveniles dispersing away from their natal ponds into forest habitat (pondside fences) or across the road (roadside fences) over two field seasons. To explain variations in abundance of dispersers among sites, we measured several road-associated variables including traffic intensity and roadside habitat structure, pond quality, and weather variables. We estimated the activity patterns (across 4-day periods) and seasonal abundance of juveniles in transit between ponds and terrestrial habitat using generalized linear mixed models. For all groups, activity across 4-day periods increased with either total precipitation or minimum air temperature. However, road-associated variables were also important for some species. Mole salamander activity was lowest next to roads. Wood frog activity increased with minimum air temperature, but the effect was weakest at roadside fences (minimum air temperature fence position interaction). Seasonal abundance of most groups varied with habitat structure or pond hydroperiod. Green frog abundance decreased with increasing traffic intensity, but abundance was higher at roadside fences than pondside fences. In contrast, wood frog seasonal abundance tended to be lowest at roadside fences. We conclude that road-associated disturbances are detectable at fine temporal scales and that amphibian responses to such variables can be influenced by weather variables. Keywords: barrier, connectivity, dispersal, frogs, salamanders, temperature, traffic, vehicles. Introduction As road networks expand with urbanization, the effects of roads and associated traffic on animal and plant populations have received attention worldwide (Findlay and Houlahan, 1997; Forman and Alexander, 1998; Forman, 2000; Underhill and Angold, 2000; Forman et al., 2003; St. Clair, 2003). Obvious effects of roads and associated infrastructures are the loss and fragmentation of habitat, as well as mortality of individuals. Mortality through collision with vehicles is well documented in mammals, birds, reptiles, amphibians, and insects (Oxley, Fen- 1 - Chaire de recherche du Canada en conservation des paysages, Département de biologie, Université de Moncton, Moncton, Nouveau-Brunswick E1A 3E9, Canada 2 - Centre d étude de la forêt, Département des sciences appliquées, Université du Québec en Abitibi- Témiscamingue, Rouyn-Noranda, Québec J9X 5E4, Canada Corresponding author; ton and Carmody, 1974; Mader, Schell and Kornacker, 1990; Fahrig et al., 1995; Reijnen, Foppen and Meeuwsen, 1996; Marchand and Litvaitis, 2004). However, for small animals such as amphibians, whose dispersal ability is constrained by humidity, paved roads themselves may represent major obstacles, even without traffic: such areas devoid of cover represent an arid substrate which may increase the vulnerability of amphibians to predation or desiccation. Roads potentially alter the movements of many animal taxa and, as a result, reduce functional connectivity (Merriam et al., 1989; Taylor et al., 1993; Shine et al., 2004; Jaeger et al., 2005; Bouchard et al., 2009). Thus, only considering road mortality may underestimate the actual effects of roads on nearby amphibian populations. During their seasonal movements, pondbreeding amphibians use different landscape elements within a few kilometers (Dole, 1971; Berven, 1990; Pope, Fahrig and Merriam, 2000; Semlitsch and Bodie, 2003). When moving between these habitat patches, amphibians often Koninklijke Brill NV, Leiden, DOI: / X625512
2 114 M. Gravel, M.J. Mazerolle, M.-A. Villard have to cross hostile environments, such as deforested areas, agricultural fields, or roads. Road presence and direct mortality caused by vehicles increase pressures on amphibian populations (Semlitsch, 2000), but indirect effects on roadside populations remain poorly known (but see Sun and Narrins, 2005; Bee and Swanson, 2007; Parris, Velik-Lord and North, 2009). Most studies evaluating the effects of roads on amphibians have focused on mortality caused by vehicles (Hodson, 1966; van Gelder, 1973; Fahrig et al., 1995; Hels and Buchwald, 2001; Mazerolle, 2004; Sutherland, Dunning and Baker, 2010) or patterns of presence or abundance at ponds in relation to road density at the landscape scale (Vos and Chardon, 1998; Carr and Fahrig, 2001; Eigenbrod, Hecnar and Fahrig, 2008), with a few recent studies aiming to predict the location of crossing hotspots (Santos et al., 2007; Glista, DeVault and DeWoody, 2008; Hartel et al., 2009; Langen, Ogden and Schwarting, 2009). Indirect effects of roads such as chemical pollution, changes in hydrological regimes, or modifications of animal behavior have recently received increasing attention (Mazerolle, Huot and Gravel, 2005; Andrews, Gibbons and Jochimsen, 2008; Karraker, 2008; Karraker, Gibbs and Vonesh, 2008). For instance, a number of studies focused on barrier effects along gravel roads in managed forests and paved roads (Gibbs, 1998a; demaynadier and Hunter, 2000; Marsh and Beckman, 2004; Marsh et al., 2005; Andrews, Gibbons and Jochimsen, 2008). However, none have yet integrated weather variables directly in their assessment of road impacts. Indeed, weather conditions can influence connectivity between habitats separated by suboptimal habitat (Chan-McLeod, 2003; Schalk and Luhring, 2010). In this study, we investigate the effects of landscape (e.g., distance to pond), anthropic variables (e.g., traffic intensity), and weather variables on roadside movement patterns of juvenile green frogs (Lithobates clamitans), wood frogs (Lithobates sylvaticus), American toads (Anaxyrus americanus), and mole salamanders (Ambystoma laterale, A. maculatum). In the present paper, we studied the effect of paved roads on the patterns of amphibian juveniles in transit from natal ponds to terrestrial habitat. To do so, we investigated activity patterns at a resolution of four days within each season as well as total abundance for each of two field seasons. We focused on juveniles as they disperse following metamorphosis (demaynadier and Hunter, 1999; Vasconcelos and Calhoun, 2004), maintain connectivity among populations (Berven and Grudzien, 1990; Cushman, 2006), are especially vulnerable to road-associated impacts due to their extensive movements, and are important determinants of population persistence (Lampo and de Leo, 1998; Vonesh and De la Cruz, 2002). We predicted a barrier effect of paved roads on small-scale movements of juveniles. Specifically, we expected that, at a given distance from the pond, juvenile numbers would be lower at drift fences separated from the pond by a road (roadside fences) than at drift fences that were not separated from the pond by a road (pondside fences). We also predicted that this effect might be exacerbated by road traffic intensity (i.e., fence position traffic intensity interaction). Based on previous studies and amphibian ecological requirements, we anticipated effects of distance to the nearest pond (Sjögren Gulve, 1994; Houlahan and Findlay, 2003), terrestrial vegetation structure (demaynadier and Hunter, 1998; Gibbs, 1998b; Marsh and Beckman, 2004), as well as pond hydroperiod and perimeter (Wellborn, Skelly and Werner, 1996; Snodgrass et al., 2000; Weyrauch and Grubb, 2004). For juvenile activity patterns within each season (i.e., every four days), we contrasted the above-mentioned hypotheses against an effect of weather variables (i.e., minimum air temperature, total precipitation) as well as interactive effects of these variables and fence position. These interactions assessed the influence of weather patterns on the effect of roads as barriers to dispersing juvenile amphibians.
3 Roads, weather, and amphibian dispersal 115 Methods Study area The study was conducted in eastern New Brunswick, in a 6400 km 2 area encompassing Kouchibouguac National Park of Canada (214 km) and its surroundings (fig. 1). The park is protected, but forestry and peat mining activities take place beyond its boundaries. Our study area has a very simple topography (lowlands of the Maritime plain) and consists of a mosaic of terrestrial habitats, mainly forests and abandoned fields, and several potential amphibian breeding habitats in the form of rivers, brooks, lakes, and ponds. Average annual rainfall is mm (Parks Canada, unpublished data). Forests (mainly mid-seral stands) are dominated by black spruce (Picea mariana), balsam fir (Abies balsamea), red maple (Acer rubrum), white birch (Betula papyrifera), and trembling aspen (Populus tremuloides). White pine (Pinus strobus) andredpine(p. resinosa) are locally common. Sampling sites Sixteen study sites were selected for this study in 2003, both within and outside the Park. Each consisted of a pond, identified as a potential breeding site, located within 70 m of a road. We selected segments of two-laned paved roads with low (<700 vehicles/day, n = 5), intermediate ( cars/day, n = 6), and high traffic intensities (>2000 cars/day, n = 5). These roads had a mean ± SD shoulderto-shoulder width of 8.6 ± 0.3 m, 9.4 ± 0.8 m, and 10.9 ± 1.4 m, for low, intermediate and high traffic roads, respectively. We selected each study site based on the following criteria: (1) similar vegetation composition on both sides of the road; and (2) the site was at least 2 km away from any other potential breeding site to reduce the chances of capturing individuals from nearby populations (Semlitsch and Bodie, 2003). In two of the sites, there were two distinct ponds separated by less than 3 m. In these cases, we assumed that these ponds represented a single population. Three sites on private lands had to be abandoned before the end of the first field season due to unexpected forestry activities by the landowners, leaving 13 sites for analyses in We replaced these three sites with three additional sites in 2004 (i.e., n = 16). Trapping design At each pond site, we used a pair of 10-m aluminium drift fences (60 cm high, cm of which was underground) associated with pitfall traps to intercept amphibians moving through the landscape (Corn and Bury, 1989). Given that we selected sites with similar vegetation on both sides of the road, we assumed that amphibians were as likely to leave the pond in either direction, and that the number of juveniles captured at each fence would be similar, unless the road acted as a barrier to movements. Fences were equidistant from the pond, on either side of the road, and positioned so that the pond-road axis was perpendicular to the road (fig. 2). In this paper, the term roadside fence denotes a fence separated from the pond by a road, whereas pondside fence refers to a fence located on the same side of the road as the pond. The distance between the pond and the pair of fences was specific to each site: it depended on the distance from the pond edge to the road. At each site, we installed the roadside fence outside the road verge and approximately 2 m into forest cover. The pondside fence was erected in continuous habitat at the same distance as the roadside fence, but on the opposite side of the pond (i.e., side facing away from the road). Thus, the distance between the pond and each fence was identical within a site, but varied among sites as a function of the distance from the pond to the road. We installed three pitfall traps on either side of each drift fence (at the ends and center), for a total of 12 pitfall traps per site. The traps (30 cm in diameter 40 cm in height) were filled up to 10 cm of water to prevent desiccation of individuals captured, and we added a sponge (ca cm) to prevent the drowning of individuals. The water level was also readjusted after rain. Finally, we added a small wooden stick in each trap to facilitate the escape of small mammals (Perkins and Hunter, 2002). Pitfall traps were opened from 26 May to 24 August 2003, and from 1 May to 24 September They were checked every 4 days. Trapping effort was identical across sites each year, with 90 and 147 trapping days for 2003 and 2004, respectively. We closed traps between trapping periods (from 24 August 2003 to 1 May 2004). Each individual captured was measured from snout to vent (SVL) and marked by toe-clipping (Donnelly and Guyer, 1994). To minimize the physiological and behavioural consequences of this procedure on the animals, we restricted marking to a single phalanx per individual per year (McCarthy and Parris, 2004). Individuals were released on the opposite side of the fence in a shady area at least 5 m from their point of capture to avoid immediate recapture and desiccation. On each visit, we recorded the number of juveniles captured for each species in each trap at a given fence. We differentiated juveniles from adults based on the mean published minimum adult snout-to-vent length from Wright and Wright (1949), Behler and King (1989), and Desroches and Rodrigue (2004). Recaptures were excluded from the analyses. Pond and site characteristics For each pond, we recorded the perimeter, pond-fence distance, and hydroperiod (calculated as the proportion of days with water over the entire trapping season). We also measured traffic intensity for the corresponding road segment as the annual average daily traffic (New Brunswick Department of Transportation and Kouchibouguac National Park of Canada), and characterized terrestrial habitat structure (see details below). These data are summarized in table 1. We quantified habitat structure in permanent 1 m quadrats around each fence. Quadrats were located to the North, South, East, and West of each fence (1 plot per direction), at two distances, i.e., one half and one quarter of the distance between the pond and fence. Again, these distances were site-specific and depended on the distance between the
4 116 M. Gravel, M.J. Mazerolle, M.-A. Villard Figure 1. Geographic location of the study area in New Brunswick, Canada.
5 Roads, weather, and amphibian dispersal 117 pond and fence. In each plot, we estimated the percent cover of the forest canopy, as well as that of woody debris. We measured canopy closure by holding a cm plexiglass grid overhead and counting the number of grid cells obscured by canopy foliage. Measurements were taken at the beginning of each month during trap operation and we used the means of monthly cover values in subsequent analyses. Weather data We used weather data from the closest weather station (Kouchibouguac National Park), obtained from the Environment Canada Climate Archives ( weatheroffice.gc.ca/). The variables considered for analysis consisted of minimum air temperature (C) and total precipitations (mm). We computed these values based on 4-day periods between visits to each site. We did not include Julian day in the analyses as it was strongly correlated with weather variables (i.e., r>0.7). Analytical strategy To address activity patterns at a resolution of 4-day periods, we modeled the counts of juveniles observed at site i, at Figure 2. Schematic illustration of a typical study site with two drift fences and associated pitfall traps at equal distances from the pond (not drawn to scale). fence j, on visit k with generalized linear mixed models with a Poisson distribution and random intercepts for sites, and fences nested within sites, respectively (Zuur et al., 2009). Specifically, we included the latter random effects to reflect the structure of our data set. In essence, random effects accounted for the correlation of observations within sites and fences nested within sites. We used the captures at each drift-fence during a given 4 day period (i.e., 6 traps open for 4 days) in the analyses, yielding two counts at each site per visit. We analyzed each species or group separately. We pooled data from both years in a single analysis and systematically included a year categorical variable in all models. The data sets consisted of 22 4-day periods in 2003 (13 sites), and 30 4-day periods in 2004 (16 sites). We specified 15 models with different effects of covariates on abundance and did not include explanatory variables that were highly correlated in the same model. These models consisted of road, pond, terrestrial habitat structure variables, weather variables, with or without interactive effects of fence position (i.e., fence separated from the pond by a road or not), based on our a priori hypotheses (see Introduction, table 2). The model set also included a null model consisting of only year and random effects. In a second analysis, we summed juvenile captures at each fence for each season and used Poisson models with a year categorical variable and random effects structure identical to that of the activity models. Because the sample size was smaller, we considered simpler models, namely (1) interactive effects of fence position and traffic intensity, (2) additive effects of fence and traffic intensity, (3) pond characteristics, (4) additive effects of pond characteristics and fence position, (5) pond characteristics and interactive effects of pond perimeter and fence position, (6) pond characteristics and interactive effects of road distance and fence position, (7) pond characteristics and interactive effects of hydroperiod and fence position, (8) canopy cover and woody debris cover, and (9) null model. Parameters in the Poisson mixed models were estimated with the Laplace approximation to the likelihood in R version using the lme4 package (Ihaka and Gentleman, 1996; Bates, Maechler and Bolker, 2010). We centered continuous covariates before analysis. We assessed the fit of the highest ranked models using a parametric bootstrap approach based on the sum of squared residuals (SSE). To Table 1. Explanatory variables included in the analysis. Explanatory variables Code Unit Range Fence position Fence Road vs Pond Pond perimeter a Perim m Pond-fence distance a Distpond m Hydroperiod Hydroperiod Proportion of trapping period during which pond contained water Canopy cover Canopy % 9-89 Woody debris cover Wood % 2-46 Minimum air temperature Min.temp C Total precipitation Total.prec mm a Natural log-transformed before analysis.
6 118 M. Gravel, M.J. Mazerolle, M.-A. Villard Table 2. Set of Poisson (or Gaussian) generalized linear mixed models estimating the number of juveniles in transit on each side of the road during 4-day periods as a function of pond, vegetation, and weather variables. All models included year as a categorical variable as well as random effects for site and fence position. Note that when Poisson models did not fit the data well, we used Gaussian models with same random effects structure. Models Simple models Road characteristics models Fence + Traffic + Fence*Traffic Fence + Traffic Pond characteristics models Perim + Distpond + Hydroperiod Pond and vegetation characteristics model Perim + Distpond + Hydroperiod + Canopy + Wood Vegetation characteristics model Canopy + Wood Weather model Min.temp + Total.prec Null model No covariates other than intercept and year Combined models with additive or interactive effects of roadside Road and pond characteristics models Fence + Traffic + Perim + Distpond + Hydroperiod Fence + Traffic + Fence*Traffic + Perim + Distpond + Hydroperiod Fence + Traffic + Perim + Fence*Perim + Distpond + Hydroperiod Fence + Traffic + Perim + Distpond + Fence*Distpond + Hydroperiod Fence + Traffic + Perim + Distpond + Hydroperiod + Fence*Hydroperiod Road and weather models Fence + Min.temp + Total.prec Fence + Min.temp + Fence*Min.temp + Total.prec Fence + Min.temp + Total.prec + Fence*Total.prec do so, we simulated 1000 data sets based on the fixed and random effects of a given model (e.g., Gelman and Hill, 2007). We then determined how often SSE from simulated data were as large or greater than the SSE of the model using the original data (P < 0.05 being evidence of lack of fit). In cases of poor model fit, we opted for Gaussian regressions with random intercepts, using the log-transformed abundance as a response variable (i.e., log(x + 1)) with the same fixed and random effects structure as in the Poisson mixed models. Gaussian models were fit by maximum likelihood and residual diagnostics were checked for deviations from normality and homoscedasticity assumptions. For the Gaussian models, the parametric bootstrap did not suggest lack of fit (P > 0.358). We ranked the models according to the AIC c and then computed model-averaged parameter estimates ( ˆ β) for the covariates in the best supported models with the AICcmodavg package (Buckland, Burnham and Augustin, 1997; Burnham and Anderson, 2002; Mazerolle, 2011). We counted each random effect as a parameter (i.e., a variance) in our model selection (Bolker et al., 2008). Results Amphibian captures in a given season varied substantially among groups: 120 mole salamanders in 2003 and 290 in 2004, 69 American toads in 2003 and 90 in 2004, 107 green frogs in 2003 and 255 in 2004, 3151 wood frogs in 2003 and 1166 in Very few individuals were recaptured (0.11% in 2003 and 2.01% in 2004). For the 4-day activity patterns of juveniles, Poisson mixed models fit well most of the data sets, with the exception of wood frogs for which we used Gaussian models with the log of abundance as a response variable (table 3). Weather models had considerably more support than the other models for all the groups studied, encompassing all the weight, but some also featured additive or interactive effects of weather and fence position. The abundance of mole salamanders increased with minimum air temperature ( ˆ β air temperature = 0.05, 95% CI: 0.03, 0.07) and total precipitation ( ˆ β total precipitation = 0.03, 95% CI: 0.02, 0.03) during our 4-day sampling intervals (fig. 3a, b). The number of American toads during 4-day periods increased with minimum air temperature ( ˆ β air temperature = 0.09, 95% CI: 0.05, 0.13), but not with total precipitation (fig. 3c). Green frog abundance increased with minimum air temperature ( ˆ β air temperature =
7 Roads, weather, and amphibian dispersal 119 Table 3. Summary of generalized linear mixed model selection based on AIC c on amphibian juvenile activity during 4-day sampling occasions across the season in response to pond, weather and road variables in eastern New Brunswick, Canada. Species Top-ranked models K AIC c Delta AIC c w i Mole salamanders Fence + Min.temp + Total.prec Fence + Min.temp + Total.prec + Fence*Total.prec Fence + Min.temp + Total.prec + Fence*Min.temp Min.temp + Total.prec American toads Min.temp + Total.prec Fence + Min.temp + Total.prec + Fence*Min.temp Fence + Min.temp + Total.prec Fence + Min.temp + Total.prec + Fence*Total.prec Green frogs Min.temp + Total.prec Fence + Min.temp + Total.prec Fence + Min.temp + Total.prec + Fence*Min.temp Fence + Min.temp + Total.prec + Fence*Total.prec Wood frogs a Fence + Min.temp + Total.prec + Fence*Min.temp Fence + Min.temp + Total.prec Min.temp + Total.prec Fence + Min.temp + Total.prec + Fence*Total.prec Notes: K denotes the number of estimated parameters including intercept, year categorical variable, and random effects, w i is the Akaike weight of the most parsimonious model. a Gaussian models with random effects were used with these data because Poisson models lacked fit. Table 4. Summary of generalized linear mixed model selection based on AIC c on the captures of amphibian juveniles summed for each season, in response to pond and road variables in eastern New Brunswick, Canada. Species Top-ranked models K AIC c Delta AIC c w i Mole salamanders a Perim + Distpond + Hydroperiod Fence + Traffic + Perim + Distpond + Hydroperiod Fence + Traffic + Perim + Fence*Perim + Distpond Hydroperiod Fence + Traffic + Perim + Distpond + Hydroperiod Fence*Hydroperiod American toads Canopy + Wood Null Fence + Traffic + Fence*Traffic Fence + Traffic Green frogs Fence + Traffic Canopy + Wood Fence + Traffic + Fence*Traffic Null Wood frogs a Null Canopy + Wood Fence + Traffic Fence + Traffic + Fence*Traffic Notes: K denotes the number of estimated parameters including intercept, year categorical variable, and random effects, w i is the Akaike weight of the most parsimonious model. a Gaussian models with random effects were used with these data because Poisson models lacked fit.
8 120 M. Gravel, M.J. Mazerolle, M.-A. Villard Figure 3. Patterns of juvenile abundance per 10 m of fence across minimum air temperature, total precipitation, and fence position for 4-day periods in 2003 and 2004 in eastern New Brunswick, Canada. Note that dotted lines indicate 95% confidence limits around the model-averaged predictions from generalized linear mixed models. Note that relationship in (c, d, e) is shown for pondside fences. 0.22, 95% CI: 0.18, 0.25) and total precipitation ( ˆ β total precipitation = 0.02, 95% CI: 0.01, 0.03; fig. 3d, e). Overall, there was weak support in favor of pond perimeter, distance to pond, canopy cover, woody cover, and hydroperiod on juvenile abundance at 4-day intervals relative to the effects of weather variables. However, road associated variables influenced mole salamander and wood frog numbers. Indeed, mole salamanders tended to be less abundant at roadside fences than at pondside fences ( ˆ β road fence = 0.58,95%CI: 1.14, 0.02;fig.3a,b).Similarly, wood frog abundance increased with minimum air temperature, but the relationship was weaker at roadside fences than pondside fences (fence position air temperature interaction,
9 Roads, weather, and amphibian dispersal 121 ˆ β road fence:air temperature = 0.02, 95% CI: 0.03, 0.01; fig. 3f). Seasonal abundance, based on the summed data, varied with habitat structure and pond hydroperiod. Again, Poisson mixed models fit well to some data sets, except for the mole salamanders and wood frogs, for which Gaussian models were used on the log of abundance. Mole salamander abundance increased with pond hydroperiod ( ˆ β hydroperiod = 1.51, 95% CI: 0.11, 2.92; fig. 4a). American toad abundance increased with canopy cover ( ˆ β canopy = 0.04, 95% CI: 0.01, 0.07; fig. 4b), whereas green frog abundance decreased with the cover of woody debris on the ground ( ˆ β woody debris = 0.05, 95% CI: 0.09, 0.02; fig. 4c). No other amphibian response variable varied with habitat variables. Certain species varied with road associated variables. Specifically, green frog seasonal abundance decreased with increasing traffic intensity ( ˆ β traffic intensity = 0.75, 95% CI: 1.35, 0.14; fig. 5a), and tended to be greater at roadside fences than pondside fences ( ˆ β road fence = 0.49, 95% CI: 0.01, 0.97; fig. 5b). In contrast, wood frog abundance was marginally lower at roadside fences than pondside fences ( ˆ β road fence = 0.43, 95% CI: 0.89, 0.03; fig. 5c). Discussion Effects of weather and roads Figure 4. Seasonal abundance, based on the sum of captures for each season in 2003 and 2004, of amphibian juveniles per 10 m of fence across habitat structure and pond hydroperiod in eastern New Brunswick, Canada. Note that dotted lines indicate 95% confidence limits around the modelaveraged predictions at pond fences from generalized linear mixed models. The activity patterns of all amphibian groups increased with either minimum air temperature or total precipitation. In two out of four cases, we found additive or interactive effects of weather and fence position. Indeed, mole salamander activity was greatest at pondside fences with additive effects of both minimum air temperature and total precipitation. Although wood frog activity increased with air temperature, the relationship was weaker at roadside fences than pondside fences and may be related to increased desiccation risk on the roads. This suggests
10 122 M. Gravel, M.J. Mazerolle, M.-A. Villard Figure 5. Seasonal abundance, based on the sum of captures for each season in 2003 and 2004, of amphibian juveniles per 10 m of fence across road variables in eastern New Brunswick, Canada. Dotted lines (a) and error bars (b, c) denote 95% confidence limits around model-averaged predictions from generalized linear mixed models. Note that relationship in (a) is shown for pond fences. that roads hinder seasonal movements between aquatic and terrestrial habitats, thereby reducing connectivity among populations. Our results highlight the need to include weather variables in models of amphibian movements, even when the study focuses on human-associated habitat disturbances such as roads or forestry. Indeed, weather patterns can influence the permeability of certain environments such as agricultural fields or clearcut plots that are otherwise hostile under suboptimal conditions (Chan-McCleod, 2003; Schalk and Luhring, 2010). In a study of seasonal ponds >300 m from paved roads, Timm, McGarigal and Compton (2007) reported that weather variables such as air temperature and amount of rainfall in the previous 24 h were good predictors of amphibian movement. However, few studies consider the mitigating effects of weather on small-scale movement patterns in amphibians. For the specific example of roads, favorable weather patterns may temporarily increase the propensity of individuals to move across roads but this can then increase road mortality. The seasonal abundance, based on the summed data, of two species varied with roadassociated variables. Green frog abundance decreased with increasing traffic intensity. However, a slightly greater number of green frogs were captured at roadside fences than at pondside fences, whereas wood frogs tended to be less abundant at roadside fences than pondside fences. These patterns could be the result of a mixture of a reluctance to cross roads or open habitat, or direct mortality from collisions with vehicles. In a study on leopard frog (Lithobates pipiens) behaviour, Bouchard et al. (2009) observed that individuals moved more slowly near roads than in areas without roads. Orlowski (2007) and Sutherland, Dunning and Baker (2010) observed a greater number of dead individuals on the roads with the lowest traffic intensities where populations were presumably higher than in areas under high traffic intensity, whereas Mazerolle (2004) reported different responses across species. The differential
11 Roads, weather, and amphibian dispersal 123 response of species relative to road variables highlights the need to conduct manipulative behavioral experiments to investigate the mechanisms regulating the patterns we observed in this study under different weather and road conditions. Pond-breeding amphibians typically migrate seasonally and factors such as forest cover influence their trajectory when moving between landscape components used for breeding, summering, and overwintering (Pope, Fahrig and Merriam, 2000; Rothermel and Semlitsch, 2002; Mazerolle and Desrochers, 2005). Recently-metamorphosed wood frogs avoid open canopy habitats (demaynadier and Hunter, 1999; Vasconcelos and Calhoun, 2004; Patrick, Hunter and Calhoun, 2006), and other pondbreeding species show similar characteristics (Gibbs, 1998a; Rothermel and Semlitsch, 2002; Chan-McLeod, 2003). Roads and associated shoulders in our study were devoid of trees and this characteristic was partially captured by the canopy cover and woody debris cover variables. We only found an effect of habitat variables in models of seasonal abundance, whereas activity patterns (across 4-day periods) were most strongly related to weather variables and fence position. Amphibian movements across certain environments, such as roads, incur physiological and behavioral costs (Rosenberg et al., 1998; Rothermel and Semlitsch, 2002; Marsh et al., 2004, 2005; Mazerolle and Desrochers, 2005). For instance, movements over pavement induce higher stress levels (i.e., corticosterone concentration in blood) than through forest habitat (Homan et al., 2003). Substrate moisture and type can also influence the ability of individuals to move across environments (Rothermel and Semlitsch, 2002; Marsh et al., 2005; Mazerolle and Desrochers, 2005). In our study, traffic-associated disturbances (light, noise, and vibration) known to disrupt amphibian behavior may have deterred individuals from crossing roads, increased the probability of mortality on the road, or both (Mazerolle, Huot and Gravel, 2005; Sun and Narrins, 2005; Bee and Swanson, 2007; Bouchard et al., 2009). Amphibian activity across 4-day periods did not vary with traffic intensity at our sites, only seasonal abundance of green frogs decreased with increasing traffic. The apparent inconsistency could stem from an indirect effect of traffic disturbance on amphibian behavior (e.g., Mazerolle, Huot and Gravel, 2005), or simply from the nature of the metric we used to quantify traffic intensity: annual average daily traffic intensity. Though this metric has been widely used in other studies (e.g., Fahrig et al., 1995; Carr and Fahrig, 2001; Eigenbrod, Hecnar and Fahrig, 2008), Mazerolle (2004) found that nightly traffic variations influence the number of amphibians dead on the road on a given night. Unfortunately, night traffic intensities were unavailable at most of our sites. Using a more refined traffic intensity metric would probably improve model performance. Conservation implications The presence of highways, railways, and urban areas in the landscape can increase the genetic isolation of amphibian populations (Reh and Seitz, 1990; Hitchings and Beebee, 1997; Vos et al., 2001; Lesbarrères, Pagano and Lodé, 2003). Our study sites mostly encompassed secondary roads and only five sites had mean annual traffic intensities greater than 2000 cars/day. Although movements across the road were limited at some sites, it seems unlikely that current populations suffer from genetic isolation because at least some individuals could successfully disperse across the roads. Marsh et al. (2008) found that wide highways hindered gene flow between terrestrial salamander populations, whereas smaller roads exerted a negligible effect. Our results suggest that habitat fragmentation generated by roads can decrease the ability of juveniles to disperse across the landscape, which in turn may reduce population persistence (Hels and Nachman, 2002; Cushman, 2006; Zanini et al., 2008). Demographic effects of fragmentation should be of special concern in
12 124 M. Gravel, M.J. Mazerolle, M.-A. Villard protected areas intersected by paved roads such as Kouchibouguac National Park. In this study, we estimated the movement patterns of juvenile amphibians in transit between natal ponds and terrestrial habitat at two temporal scales. An analysis of seasonal abundance revealed effects of road-associated variables on two amphibian species. At a finer resolution of 4-days across the entire season, results showed that weather variables can be good predictors of amphibian activity, but also suggested additive or interactive effects of road-associated variables and weather variables. Studies conducted under a limited range of weather conditions may misrepresent the barrier effect presented by roads. This highlights the importance of considering weather variables when investigating the response of amphibians to anthropic disturbances. Acknowledgements. M.-A.Guitard,M.Huot,M.Prévost and P.-E. Hébert contributed to field work, É. Tremblay provided logistical support. G. Thompson shared traffic intensity data, and A. Beaudet created the study area maps. A. Chiasson, D. Lesbarrères, S. Reebs, and P. Maltais, and two reviewers provided constructive comments. The protocol was approved by Université de Moncton s Animal Care Committee and provincial conservation guidelines were strictly followed. Financial support was provided by the New Brunswick Wildlife Trust Fund and by a discovery grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) to MAV, and NSERC postdoctoral fellowship to MJM. This work would not have been possible without the logistical support of Kouchibouguac National Park of Canada. References Andrews, K.M., Gibbons, J.W., Jochimsen, D.M. (2008): Ecological effects of roads on amphibians and reptiles: a literature review. In: Urban Herpetology, p Mitchell, J.C., Jung Brown, R.E., Bartholomew, B., Eds, Salt Lake City, Society for the Study of Amphibians and Reptiles. Bates, D., Maechler, M., Bolker, B. (2010): lme4: Linear mixed-effects models using S4 classes. R package version ( lme4). Bee, M.A., Swanson, E.M. (2007): Auditory masking of anuran advertisement calls by road traffic noise. Anim. Behav. 74: Behler, J.L., King, W.F. (1989): National Audubon Society Field Guide to Reptiles and Amphibians. New York, Alfred A. Knopf. Berven, K.A. (1990): Factors affecting population fluctuations in larval and adult stages of the wood frog (Rana sylvatica). Ecology 71: Berven, K.A., Grudzien, T.A. (1990): Dispersal in the wood frog (Rana sylvatica): implications for genetic population structure. Evolution 44: Bolker, B.M., Brooks, M.E., Clark, C.J., Geatrendsge, S.W., Poulsen, J.R., Stevens, M.H.H., White, J.-S.S. (2008): Generalized linear mixed models: a practical guide for ecology and evolution. Trends Ecol. Evol. 24: 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: 23 ( ecologyandsociety.org/vol14/iss2/art23/). Buckland, S.T., Burnham, K.P., Augustin, N.H. (1997): Model selection: an integral part of inference. Biometrics 53: Burnham, K.P., Anderson, D.R. (2002): Model Selection and Multimodel Inference: a Practical Informationtheoretic Approach, 2nd Edition. New York, Springer- Verlag. Carr, L.W., Fahrig, L. (2001): Effect of road traffic on two amphibian species of differing vagility. Conserv. Biol. 15: Chan-McLeod, A.C.A. (2003): Factors affecting the permeability of clearcuts to red-legged frogs. J. Wildl. Manage. 67: Corn, P.S., Bury, R.B. (1989): Logging in Western Oregon: responses of headwater habitats and stream amphibians. For. Ecol. Manage. 29: Cushman, S.A. (2006): Effects of habitat loss and fragmentation on amphibians: a review and prospectus. Biol. Conserv. 128: demaynadier, P.G., Hunter, M.L., Jr. (1998): Effects of silvicultural edges on the distribution and abundance of amphibians in Maine. Conserv. Biol. 12: demaynadier, P.G., Hunter, M.L., Jr. (1999): Forest canopy closure and juvenile emigration by pool-breeding amphibians in Maine. J. Wildl. Manage. 63: demaynadier, P.G., Hunter, M.L., Jr. (2000): Road effects on amphibian movements in a forested landscape. Nat. Areas J. 20: Desroches, J.-F., Rodrigue, D. (2004): Amphibiens et Reptiles du Québec et des Maritimes. Waterloo, Éditions Michel Quintin. Dole, J.W. (1971): Dispersal of recently metamorphosed leopard frogs, Rana pipiens. Copeia 1971: Donnelly, M.A., Guyer, C. (1994): Mark-recapture. In: Measuring and Monitoring Biological Diversity: Standard Methods for Amphibians, p Heyer, W.R., McDiarmid, R.W., Hayek, L.-A.C., Foster, M.S., Eds, Washington, DC, Smithsonian Institution Press. Eigenbrod, F., Hecnar, S.J., Fahrig, L. (2008): The relative effects of road traffic and forest cover on anuran populations. Biol. Conserv. 141:
13 Roads, weather, and amphibian dispersal 125 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: Findlay, S.C., Houlahan, J. (1997): Anthropogenic correlates of species richness in southeastern Ontario wetlands. Conserv. Biol. 11: Forman, R.T.T. (2000): Estimate of the area affected ecologically by the road system in the United States. Conserv. Biol. 14: Forman, R.T.T., Alexander, L.E. (1998): Roads and their major ecological effects. Annu. Rev. Ecol. Syst. 29: Forman, R.T.T., Sperling, D., Bissonette, J.A., Clevenger, A.P., Cutshall, C.D., Dale, V.H., Fahrig, L., France, R., Goldman, C.R., Heanue, K., Jones, J.A., Swanson, F.J., Turrentine, T., Winter, T.C. (2003): Road Ecology: Science and Solutions. Washington, DC, Island Press. Gelman, A., Hill, J. (2007): Data Analysis Using Regression and Multilevel/Hierarchical Models. New York, Cambridge University Press. Gibbs, J.P. (1998a): Amphibian movements in response to forest edges, roads, and streambeds in southern New England. J. Wildl. Manage. 62: Gibbs, J.P. (1998b): Distribution of woodland amphibians along a forest fragmentation gradient. Landsc. Ecol. 13: Glista, D.J., DeVault, T.L., DeWoody, J.A. (2008): Vertebrate road mortality predominantly impacts amphibians. Herpetol. Conserv. Biol. 3: Hartel, T., Moga, C.I., Öllerer, K., Puky, M. (2009): Spatial and temporal distribution of amphibian road mortality with a Rana dalmatina and Bufo bufo predominance along the middle section of the Târnava Mare basin, Romania. North-West. J. Zool. 5: Hels, T., Buchwald, E. (2001): The effect of road kills on amphibian populations. Biol. Conserv. 99: Hels, T., Nachman, G. (2002): Simulating viability of a spadefoot toad Pelobates fuscus metapopulation in a landscape fragmented by a road. Ecography 25: Hitchings, S.P., Beebee, T.J.C. (1997): Genetic substructuring as a result of barriers to gene flow in urban Rana temporaria (common frog) populations: implications for biodiversity conservation. Heredity 79: Hodson, N.L. (1966): A survey of road mortality in mammals (and including data for the grass snake and common frog). J. Zool. 148: Homan, R.N., Regosin, J.V., Rodrigues, D.M., Reed, J.M., Windmiller, B.S., Romero, L.M. (2003): Impacts of varying habitat quality on the physiological stress of spotted salamanders (Ambystoma maculatum). Anim. Conserv. 6: Houlahan, J.E., Findlay, C.S. (2003): The effects of adjacent land use on wetland amphibian species richness and community composition. Can. J. Fish. Aquat. Sci. 60: Ihaka, R., Gentleman, R. (1996): R: a language for data analysis and graphics. J. Comput. Graph. Stat. 5: Jaeger, J.A.G., Bowman, J., Brennan, J., Fahrig, L., Bert, D., Bouchard, J., Charbonneau, N., Frank, K., Gruber, B., von Toschanowitz, K.T. (2005): Predicting when animal populations are at risk from roads: an interactive model of road avoidance behavior. Ecol. Model. 185: Karraker, N.E. (2008): Impacts of road deicing salts on amphibians and their habitats. In: Urban Herpetology, p Mitchell, J.C., Jung Brown, R.E., Bartholomew, B., Eds, Salt Lake City, Society for the Study of Amphibians and Reptiles. Karraker, N.E., Gibbs, J.P., Vonesh, J.R. (2008): Impacts of road deicing salt on the demography of vernal poolbreeding amphibians. Ecol. Appl. 18: Lampo, M., de Leo, G.A. (2011): The invasion ecology of the toad Bufo marinus: from South America to Australia. Ecol. Appl. 8: Langen, T.A., Ogden, K.M., Schwarting, L.L. (2009): Predicting hot spots of herpetofauna road mortality along highway networks. J. Wildl. Manage. 73: Lesbarrères, D., Pagano, A., Lodé, T. (2003): Inbreeding and road effect zone in a Ranidae: the case of Agile frog, Rana dalmatina Bonaparte, C. R. Biologies 326: S68-S72. Mader, H.-J., Schell, C., Kornacker, P. (1990): Linear barriers to arthropod movements in the landscape. Biol. Conserv. 54: Marchand, M.N., Litvaitis, J.A. (2004): Effects of habitat features and landscape composition on the population structure of a common aquatic turtle in a region undergoing rapid development. Conserv. Biol. 18: Marsh, D.M., Beckman, N.G. (2004): Effects of forest roads on the abundance and activity of terrestrial salamanders. Ecol. Appl. 14: Marsh, D.M., Thakur, K.A., Bulka, K.C., Clarke, L.B. (2004): Dispersal and colonization through open fields by a terrestrial, woodland salamander. Ecology 85: Marsh, D.M., Milam, G.S., Gorham, N.P., Beckman, N.G. (2005): Forest roads as partial barriers to terrestrial salamander movement. Conserv. Biol. 19: Marsh, D.M., Page, R.B., Hanlon, T.J., Corritone, R., Little, E.C., Seifert, D.E., Cabe, P.R. (2008): Effects of roads on patterns of genetic differentiation in red-backed salamanders, Plethodon cinereus. Conserv. Genet. 9: Mazerolle, M.J. (2004): Amphibian road mortality in response to nightly variations in traffic intensity. Herpetologica 60: Mazerolle, M.J. (2011): AICcmodavg: Model selection and multimodel inference based on (Q)AIC(c). R package version 1.17 ( AICcmodavg). Mazerolle, M.J., Desrochers, A. (2005): Landscape resistance to frog movements. Can. J. Zool. 83: Mazerolle, M.J., Huot, M., Gravel, M. (2005): Behavior of amphibians on the road in response to car traffic. Herpetologica 61: McCarthy, M.A., Parris, K.M. (2004): Clarifying the effect of toe clipping on frogs with Bayesian statistics. J. Appl. Ecol. 41:
14 126 M. Gravel, M.J. Mazerolle, M.-A. Villard Merriam, G., Kozakiewicz, M., Tsuchiya, E., Hawley, K. (1989): Barriers as boundaries for metapopulations and demes of Peromyscus leucopus in farm landscapes. Landsc. Ecol. 2: 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: Oxley, D.J., Fenton, M.B., Carmody, G.R. (1974): The effects of roads on populations of small mammals. J. Appl. Ecol. 11: Parris, K.M., Velik-Lord, M., North, J.M.A. (2009): Frogs call at a higher pitch in traffic noise. Ecol. Soc. 14: 25 ( Patrick, D.A., Hunter, M.L., Jr., Calhoun, A.J.K. (2006): Effects of experimental forestry treatments on a Maine amphibian community. For. Ecol. Manage. 234: Perkins, D.W., Hunter, M.L., Jr. (2002): Effects of placing sticks in pitfall traps on amphibian and small mammal capture rates. Herpetol. Rev. 33: Pope, S.E., Fahrig, L., Merriam, H.G. (2000): Landscape complementation and metapopulation effects on leopard frog populations. Ecology 81: Reh, W., Seitz, A. (1990): The influence of land use on the genetic structure of populations of the common frog Rana temporaria. Biol. Conserv. 54: Reijnen, R., Foppen, R., Meeuwsen, H. (1996): The effects of traffic on the density of breeding birds in Dutch agricultural grasslands. Biol. Conserv. 75: Rosenberg, D.K., Noon, B.R., Megahan, J.W., Meslow, E.C. (1998): Compensatory behavior of Ensatina eschscholtzii in biological corridors: a field experiment. Can. J. Zool. 76: Rothermel, B.B., Semlitsch, R.D. (2002): An experimental investigation of landscape resistance of forest versus old-field habitats to emigrating juvenile amphibians. Conserv. Biol. 16: Santos, X., Llorente, G.A., Montori, A., Carretero, M.A., Franch, M., Garriga, N., Richter-Boix, A. (2007): Evaluating factors affecting amphibian mortality on roads: the case of the common toad Bufo bufo, near a breeding place. Anim. Biodivers. Conserv. 30: Schalk, C.M., Luhring, T.M. (2010): Vagility of aquatic salamanders: implications for wetland connectivity. J. Herpetol. 44: Semlitsch, R.D. (2000): Principles for management of aquatic-breeding amphibians. J. Wildl. Manage. 64: Semlitsch, R.D., Bodie, J.R. (2003): Biological criteria for buffer zones around wetlands and riparian habitats for amphibians and reptiles. Conserv. Biol. 17: Shine, R., LeMaster, M.P., Wall, M., Langkilde, T., Mason, R.T. (2004): Why did the snake cross the road? Effects of roads on movement and location of mates by garter snakes (Thamnophis sirtalis parietalis). Ecol. Soc. 9: 9 ( Sjögren Gulve, P. (1994): Distribution and extinction patterns within a northern metapopulation of the pool frog, Rana lessonae. Ecology 75: Snodgrass, J.W., Komoroski, M.J., Bryan, A.L., Burger, J. (2000): Relationships among isolated wetland size, hydroperiod, and amphibian species richness: implications for wetland regulations. Conserv. Biol. 14: St. Clair, C.C. (2003): Comparative permeability of roads, rivers, and meadows to songbirds of Banff National Park. Conserv. Biol. 17: Sun, J.W.C., Narins, P.M. (2005): Anthropogenic sounds differentially affect amphibian call rate. Biol. Conserv. 121: Sutherland, R.W., Dunning, P.R., Baker, W.M. (2010): Amphibian encounter rates on roads with different amounts of traffic and urbanization. Conserv. Biol. 24: Taylor, P.D., Fahrig, L., Henein, K., Merriam, G. (1993): Connectivity is a vital element of landscape structure. Oikos 68: Timm, B.C., McGarigal, K., Compton, B.W. (2007): Timing of large movement events of pond-breeding amphibians in western Massachussetts, USA. Biol. Conserv. 136: Underhill, J.E., Angold, P.E. (2000): Effects of roads on wildlife in an intensively modified landscape. Environ. Rev. 8: van Gelder, J.J. (1973): A quantitative approach to the mortality resulting from traffic in a population of Bufo bufo L. Oecologia 13: Vasconcelos, D., Calhoun, A.J.K. (2004): Movement patterns of adult and juvenile Rana sylvatica (LeConte) and Ambystoma maculatum (Shaw) in three restored seasonal pools in Maine. J. Herpetol. 38: Vonesh, J.R., De la Cruz, O. (2002): Complex life cycles and density dependence: assessing the contribution of egg mortality to amphibian declines. Oecologia 133: Vos, C.C., Chardon, J.P. (1998): Effects of habitat fragmentation and road density on the distribution pattern of the moor frog Rana arvalis. J.Appl.Ecol.35: Vos, C.C., Antonisse-De Jong, A.G., Goedhart, P.W., Smulders, M.J.M. (2001): Genetic similarity as a measure for connectivity between fragmented populations of the moor frog (Rana arvalis). Heredity 86: Wellborn, G.A., Skelly, D.K., Werner, E.E. (1996): Mechanisms creating community structure across a freshwater habitat gradient. Annu. Rev. Ecol. Syst. 27: Weyrauch, S.L., Grubb, T.C., Jr. (2004): Patch and landscape characteristics associated with the distribution of woodland amphibians in an agricultural fragmented landscape: an information-theoretic approach. Biol. Conserv. 115: Wright, A.H., Wright, A.A. (1949): Handbook of Frogs and Toads of the United States and Canada, 3rd Edition. Ithaca, Comstock Publishing Company.