Thrice as easy to catch! Copper and temperature modulate predator‐prey interactions in larval dragonflies and anurans

Abstract

Amphibians are important indicators of environmental health, and their populations are in worldwide decline. The causes of these declines are diverse and not well understood. In some cases multiple stressors and complex causal mechanisms have been identified. Experimental studies have shown that contaminants can cause the failure of Lithobates sylvaticus tadpoles to initiate predator avoidance behaviors, potentially leading to increased tadpole capture and injury. Copper is a contaminant known to negatively affect amphibians and other aquatic organisms at sub‐lethal levels. Mining waste, certain pesticides, vehicle exhaust and brake pad dust are sources of copper, which can enter hydrologic systems through runoff. Additionally, temperature is known to influence predator‐prey interactions of ectotherms and is predicted to rise in some areas as climate changes. We examined how copper and temperature affected behavior and predation dynamics between an odonate predator (Aeshna sitchensis) and larval L. sylvaticus prey. We found that sublethal concentrations of copper near the analytical detection limits for this element (1.85 μg Cu/L) significantly reduced tadpole and odonate activity. Above‐average temperatures (22°C) significantly increased tadpole activity and decreased dragonfly activity, compared with ambient‐temperature treatments (17°C). These behavioral responses culminated in an approximately three‐fold increase in the number of dragonfly attacks on tadpoles in the elevated‐temperature, copper‐exposed treatments. We suggest that increased concentrations of dissolved copper and elevated water temperatures are harmful to amphibian prey through maladaptive behavioral responses in the presence of predators.

Introduction

Amphibians are excellent indicators of environmental quality because their permeable skin allows for exchange of materials with the surrounding environment (Quaranta et al. 2009). The combination of multiple stressors, such as rising temperatures, environmental degradation, and shifting ecosystem dynamics, can lead to greater stress and negative consequences for amphibian populations. Climate change is expected to be the principal direct driver of altered ecosystem services and biodiversity loss (Millennium Ecosystem Assessment 2005). Consequences of multiple stressors include behavioral changes and increased prey injuries and fatalities. Nevertheless, combinations of multiple stressors can produce unforeseen consequences for individual organisms and natural communities (Relyea 2010). Aquatic ecosystems contain a large number of threatened species (Wilcove et al. 2000), and are particularly susceptible to pollutant concentration due to terrestrial runoff (Scher and Thiery 2005, Snodgrass et al. 2008). As such, it is especially important to focus on multiple stressors to organisms in these habitats. Previous studies have shown that low levels of contaminants in aquatic environments can significantly affect many aspects of an organism's physiology (e.g., neurotransmitters, hormones, immune responses), reproduction, morphology and behavior (Relyea and Hoverman 2006, Bridges 1999a, b, Broomhall 2002). In addition to the direct effects on individual organisms, it is important to understand how multiple stressors alter species interactions. Understanding predation dynamics requires increased knowledge about the effects of multiple stressors in aquatic systems (Johnson and Bowerman 2010). Additionally, knowing how multiple stressors change community dynamics may contribute to understanding the decline of amphibian populations.

Global amphibian population declines and extinctions have been documented over the past few decades (Wake 1991, Stuart et al. 2004, Hoffman et al. 2010). Local disappearances have also been reported in the available long‐term records (Blaustein et al. 1994). Even seemingly pristine habitats have experienced declines, with 33% of amphibian species globally threatened, and up to 41% experiencing population declines (Stuart et al. 2004, Hoffman et al. 2010). Meanwhile, only 0.5% of populations were increasing in size (Stuart et al. 2004). The causes of declines are varied, with some related to global climate change and observed shifts in ecosystems and others related to disease (Whiles et al. 2012). Range‐restricted species are extremely sensitive to shifts in climate, such as alteration of normal temperature and precipitation patterns (Sodhi et al. 2008). These sensitivities may be exacerbated by species life history traits (e.g., the tendency to reproduce in ephemeral habitats), small population sizes, short generation times, or fewer occupied areas (Pearson et al. 2014). Climatic variability, pollution, and habitat isolations are all increasing threats to amphibian populations (Sodhi et al. 2008, Brühl et al. 2013). Recent climatic shifts have caused some range‐restricted species to go extinct, and amphibians are among those experiencing the most detrimental effects of climate change (Parmesan 2006). Decreasing population sizes can also be attributed to human impacts and developments. Environmentally relevant levels of pesticide exposure have been shown to cause acute mortality in terrestrial life stages of European common frogs (Brühl et al. 2013). In a recent study, juvenile European common frogs were subjected to just a few of the several thousand registered pesticides at government‐accepted levels and after 7 days mortality was between 40 and 100% (Brühl et al. 2013).

Contaminants are thought to be a potential culprit in the decline of amphibian populations, specifically through their effects on amphibian behaviors relating to foraging and predator‐prey interactions in the breeding habitat (Relyea 2010). Entire bodies of toxicological work have focused on the effects of contaminants on individual species (Moriarty 1999, Sparling et al. 2010), yet recently, a variety of studies have focused on the potential indirect or community level effects of sub‐lethal concentrations of contaminants. For example, copper‐intoxicated tadpoles have demonstrated maladaptive behaviors, which may make them more vulnerable to predatory attacks (Reeves et al. 2011). Tadpoles exposed to toxicants have failed to utilize refugia in the presence of a predator, and spent more time in refugia in the absence of a predator, a behavior that decreases feeding opportunities, potentially limiting growth (Bridges 1999a). Additionally, toxicants may become more lethal when abiotic stressors like temperature or drought stress are added to biotic stressors, like parasites, predators, or diseases (Relyea 2004, 2005, Relyea and Diecks 2008, Rohr et al. 2008).

Copper is a contaminant known to have a negative effect on amphibians and other aquatic organisms at sub‐lethal levels (Sandahl et al. 2007, Reeves et al. 2011). Copper is toxic to aquatic organisms and is found in both road runoff and mining waste. Vehicle exhaust and brake‐pad dust are prominent sources of copper, which can runoff to aquatic systems (Sansalone and Buchberger 1997). Experimental studies have shown that copper, even at extremely low concentrations, approaching the limits of analytical detection (2 μg Cu/L), inhibited the olfactory system in juvenile Coho salmon and impeded predator avoidance behaviors (Sandahl et al. 2007). Reeves et al. (2011) found that wood frog tadpoles (Rana sylvatica, now known as L. sylvaticus) in aquaria treated with low copper concentrations (5 μg Cu/L) spent significantly more time at the surface of the aquaria, a maladaptive behavior in the presence of predators because tadpoles at the water surface may be more easily seen and captured by dragonfly predators. In addition, they found that copper treatment halved tadpole movement frequency, and addition of predator cues reduced tadpole activity further, suggesting a mechanism for increased predation of tadpoles through reduced foraging and slower growth in the presence of both stressors.

The Kenai National Wildlife Refuge (NWR), Alaska, from which tadpoles were sampled for this study, has 345 km of roads, which provide a potential pathway for copper and other contaminants to drain into basins and catchments (Reeves et al. 2010). Field observations and sampling have shown copper to be present in the Kenai NWR, and copper was one of several contaminants correlated with the presence of skeletal abnormalities (Reeves et al. 2010, 2011). In 2010, six wetlands sampled repeatedly in the Kenai NWR had a maximum copper concentration of 7.85 μg Cu/L and a mean concentration of 2.13 μg Cu/L (Reeves et al. 2011, doi:10.5061/dryad.sq72d).

Water temperature also plays an important role in amphibian behavior and predator‐prey relationships involving ectotherms, which control body temperature through outside heat sources (Anderson et al. 2001). To evaluate the effect of temperature on aquatic ectotherms, Anderson et al. (2001) observed tadpole behavior at three different temperatures: 9.9°C, 20.7°C and 25.7°C. They produced a model based on their findings, predicting tadpoles raised in a warm water treatment would have a higher probability of capture than tadpoles raised in a cool water treatment. In an experimental study of R. clamitans tadpoles, Moore and Townsend (1998) also determined that increased temperatures led to both increased activity and time spent at the surface, behaviors often associated with higher predation rates. To our knowledge, there are no studies examining how L. sylvaticus tadpoles respond to predation pressure at different temperatures (and no amphibian studies have experimentally investigated the combination of temperature and elevated copper concentrations). Current climate models predict global changes in temperature, which will have significant effects on hydrological systems in Southcentral Alaska, such as wetland shrinkage and decreases in water balance already documented on the Kenai NWR (Berg et al. 2009). These water balance decreases and increased temperatures have already led to rising tree lines, drying of soils and ponds, and encroachment of forest habitats into former wetland areas in the Kenai Peninsula (Klein et al. 2005). Other wetland changes between 1951 and 1996 on the Kenai Peninsula include a 66% decrease in herbaceous cover, 70% increase in shrubs, 160% increase in open woodlands, and 25% increase in closed canopy forests (Berg et al. 2009). If L. sylvaticus is adversely affected by increasing temperatures, this species will experience even greater stress than is currently being placed upon it, by habitat fragmentation by roads and contaminants documented in prior investigations (Reeves et al. 2010).

Contaminant‐induced maladaptive behaviors may have negative consequences for tadpole populations. Predators and competitors both cause larval amphibians to alter their behavior (Relyea 2001, Van Buskirk 2009). Behavioral effects caused by multiple stressors also have the ability to alter species interactions (e.g., Preisser et al. 2005). Many studies have focused on pesticides and the potential impacts on organisms, often times in the presence of a predator (Relyea 2010). In these studies, prey have responded to the presence of a predator and contaminant together by altering activity levels, decreasing feedings, failing to seek refuge, and experiencing inability to detect a predator (Broomhall 2004, Bridges 1999a, b, Relyea and Edwards 2010, Reeves et al. 2011). In several studies, lethal predation was a common outcome of these behavioral changes (Bridges 1999b, Broomhall 2002, 2004, Relyea and Edwards 2010). Understanding how amphibians behave in the presence of a predator, contaminants, and altered temperatures will be key to developing management strategies for declining amphibian populations.

The purpose of this study was to determine the effects of extremely low and field‐relevant concentrations of copper and slightly elevated temperatures on the predation dynamics between a larval aeshnid (dragonfly) predator and its larval amphibian prey. This study was designed to answer the question, how do two potential stressors, temperature and copper, interact to alter predation risk for tadpoles challenged with a macroinvertebrate predator? Findings from this study will help us understand how L. sylvaticus tadpoles may be affected by the combination of a changing climate and road‐based human disturbance.

Methods

Sampling

Wood frog egg masses were collected from four ponds on the Kenai Peninsula in southcentral Alaska during May 7–10, 2012, and transported to Anchorage. These four ponds were part of an ongoing field study assessing amphibian malformations in this area. Egg masses were sampled from these four sites by collecting a subset of no more than 10% of the mass, taking care not to deplete any single pond or individual egg mass, as well as to avoid areas where the disease‐producing fungus, Batrachochytrium dendrobatidis, had been detected (doi:10.5061/dryad.sq72d). Four or five sub‐samples of approximately 10% of the egg mass were taken from each pond, resulting in the collection of roughly 1000 eggs. Water was also collected from each pond for use in hatching and rearing. Additionally, approximately 60 odonate larvae (Aeshna sitchensis) were transported from multiple study sites in the Kenai NWR to Anchorage during June and July.

Kenai NWR wetlands sampled over three field seasons resulted in a maximum analytical detection limit for copper of 1.85 μg/L, moreover this was the mean value of 426 water samples collected from 36 study sites at Kenai between 2010 and 2012 (doi:10.5061/dryad.sq72d). Field data from study sites in the Kenai NWR suggest water temperatures averaging between 13 and 18°C during tadpole development (doi:10.5061/dryad.sq72d). Two treatment temperatures were chosen based on these field data: one simulating current average conditions at field study sites, 17°C, and a slightly elevated temperature treatment of 22°C, simulating a potential future climate (Chapman and Walsh 2007). See Appendix for further methodology.

Animal husbandry

After transportation to Anchorage, egg masses were allowed to develop together in 11.4 L glass aquaria: one aquarium for each of the four sites. Initially, egg masses from each site were kept separate to avoid potential disease‐based mortality. To avoid abrupt transition, each tank was filled with 3 L of water, comprising 1.5 L of site water and 1.5 L of clean water (Alaska's Best Water, Anchorage, AK, USA). After one week all egg masses appeared to be healthy and developing normally, so egg masses were randomly assigned to aquaria to remove potential bias due to site effects. At this point, all aquaria were filled with Alaska's Best Water. Due to slower than anticipated tadpole development outdoors, tadpoles were randomly mixed into three 75.7‐L plastic tanks and one 18.9‐L glass aquarium and transported to an indoor lab 43 days after collection, at which point water temperature was increased (average 22°C). Tadpoles were fed Zoo Med Aquatic Frog and Tadpole Food (Zoo Med Laboratories, San Luis Obispo, CA, USA) on a daily basis and supplemented with fresh organic spinach.

Prior to and through the initial trials of the experiment (trials 1–4), aeshnids were kept in a refrigerator (3°C) with a glass door to receive light, but at a cool temperature to maintain a torpor‐like state. Each aeshnid was housed in its own container to avoid intraspecific attacks, and fed weekly with insects from local ponds. Forty‐eight hours prior to a trial, aeshnids were taken out of the refrigerator and all food sources were removed. During later trials (5–13), aeshnids were not kept in the refrigerator but left at room temperature to raise their metabolic rates, increase their appetites and encourage more species interaction during trials. The weekly feedings and 48‐hour fasting periods were maintained through the remainder of the trial periods.

Experimental design

This experiment was a 2 × 2 × 2 factorial design testing the effects of copper (0 and 1.85 μg/L Cu²⁺ added as CuCl₂), temperature (17°C and 22°C), and a predator (presence or absence) on tadpole behavior and dynamics of predation (Table 1). Experimental copper concentrations were determined based on prior experimental work and measured field concentrations (Reeves et al. 2011, doi:10.5061/dryad.sq72d). A total of 13 trials were conducted, with 8 treatments each: (1) 17°C, no copper, with predator (17 −Cu +Pred); (2) 17°C, no copper, no predator (17 –Cu −Pred); (3) 17°C, with copper, with predator (17 +Cu +Pred); (4) 17°C, with copper, no predator (17 +Cu −Pred); (5) 22°C, no copper, with predator (22 –Cu +Pred); (6) 22°C, no copper, no predator (22 –Cu −Pred); (7) 22°C, with copper, with predator (22 +Cu +Pred); (8) 22°C, with copper, no predator (22 +Cu −Pred). We conducted 13 trials, each including these eight treatments in a completely randomized design. The experiment consisted of trials repeated from July 11 to August 14, 2012. Each trial used 5 new tadpoles per tank and one new dragonfly larvae for a total of 520 tadpoles and 52 dragonfly larvae (dragonflies were only present in half the treatments).

Summary of model parameters used in the best‐fit model for total number of attacks.

Experimental aquaria and set‐up

Each treatment was assigned to one of eight 9.46‐L aquaria, measuring 19.7 × 28.9 × 35.6 cm (Aqueon Aquarium Mini‐Bow 2.5 Gallon Acrylic Aquarium, Franklin, WI, USA), which were modified with a grated partition dividing each aquarium in half. A CoolWorks IceProbe Thermoelectric Aquarium Chiller (IPAC‐50W; CoolWorks, San Rafael, CA, USA) was attached to individual tanks to maintain cool temperatures; the warmer temperature aquaria were left at room temperature. Treatments targeting the cooler temperature of 17°C were measured to be within a range of 15.5–17.3°C during all trials. Treatments targeting the elevated temperature of 22°C were recorded to be within the range of 19.4–22.8°C. Aquaria that received the copper and predators were dedicated as such for all trials to avoid cross‐contamination. Alaska's Best Water (Anchorage, AK, USA) was used in all treatments. Water quality parameters were measured prior to the first trial using a YSI multi parameter sonde, and the following ranges of values recorded: specific conductivity, 9–10 μS/cm; salinity, 0 ppt; pH, 6.29–6.58; dissolved oxygen, 9.11–9.36 mg/L. Copper concentrations were measured at <0.38 μg/L for the control treatment and 1.34 μg/L for the copper treatment and water hardness was below detection limits for Ca and Mg, as it is at several high malformation study sites in Kenai (doi:10.5061/dryad.sq72d). Between each trial, aquaria were rinsed with water from the same source used for the experiments. Each aquarium was filled only halfway with water, with the goal of limiting the space available in the aquaria to increase species interactions. White paper covered three walls of the container so the tadpoles and dragonfly larvae would not be influenced by outside movements. The dragonflies attached themselves to the partition in each container; therefore, adding sticks or string as attachment points was not necessary. In an effort to control for potential covariates, all tadpoles and dragonfly larvae were measured and assessed for injuries prior to the start of the experiment. Only uninjured tadpoles and larvae were used in experiments. Following the experimental methods of Reeves et al. (2011), all trials were conducted when tadpoles were between the developmental stages 26–41 (Gosner 1960), from limb emergence until the beginning of metamorphosis. Once they reached the proper developmental stage, five tadpoles were placed in one side of each 9.46‐L container for the duration of the trial, divided in half by a grated partition. One larval predator was added to the other half of the container in the appropriate treatments. Both species acclimated to these surroundings for 24 hours before each trial began. After the acclimation period, the tadpoles and dragonfly larvae were placed together in one half of the aquarium.

Measurement of variables

Each trial lasted two hours, during which observations of tadpole and dragonfly behavior were made and simultaneously recorded with a digital video camera. Two digital cameras were utilized, each recording two tanks at a time. The following variables were recorded: total number of attacks made in a tank, tadpole position (top or bottom half of container), tadpole behavior (moving or still), and dragonfly larvae behavior (moving or still). Attacks were further classified as those resulting in the death of a tadpole (fatal attack) and those that did not kill the tadpole but during which a dragonfly attempted an attack (nonfatal attack). Video recording was continuous for 2 hours, but observations of organism movement and position were recorded at 1‐minute intervals, yielding a total of 120 point observations for each treatment. Measurements were used to determine: (1) number of times a dragonfly lunged at or made contact with a tadpole, (2) time of all attacks or attempts, (3) the number of attacks resulting in tadpole injury or death, (4) tadpole activity levels, (5) tadpole vertical position, and (6) dragonfly vertical position.

Data analysis

Statistical analyses were conducted using generalized linear mixed effect regression models (GLMMs), which allow for the estimation of fixed effects and the inclusion of random effects to control for autocorrelation (R version 2.15.2; lmer command in lme4 package, version 0.999999‐2; lsmeans command in lsmeans package; pvals.fnc function in LanguageR package; Bates et al. 2013). All models used trial as a random effect. Distributional assumptions matched the response variables, which differed across analyses, as described below.

We asked questions about predation dynamics and larval anuran and odonate behavior using a series of GLMMs, some of which required different input data sets. A separate analysis was conducted for each of the following six response variables: Total Attacks, Fatal Attacks, Time to First Attack, Tadpole Activity, Tadpole Position, and Dragonfly Activity. In these analyses we tested the effects of copper, elevated temperature, and in some cases predator presence or predator behavior, depending on the response. We used the following covariates: developmental stage (Gosner 1960), dragonfly length (mm), and whether a fatal attack occurred in a tank, because we found a fatal attack of a tadpole during a trial to be an important driver of tadpole and dragonfly behavior, a posteriori. We used Akaike's Information Criterion (AICc) to compare candidate models and determine whether the inclusion of interactions and covariates was warranted (Burnham and Anderson 2002). If AICc values differed by less than 2 then the more parsimonious model was used. The different analyses and their input datasets are described below.

Total attacks

To test the potential for mortality or injury combined as a response, as well as the influence of the main effects on the propensity of dragonflies to attack tadpoles, we used the sum of fatal and non‐fatal attacks (i.e., the number of times a dragonfly larvae lunged at or tried to capture a tadpole, regardless of capture success) as a response variable. We included all treatments that included a predator for the input dataset (i.e., four of the eight experimental treatments in each trial). We tested copper (with copper or without) and temperature (ambient or elevated) as fixed effects and dragonfly size and tadpole developmental stage as covariates. Because our response variable was a count of the total number of attacks per tank per trial, we used a Poisson distribution to model these results.

Fatal attacks

To test the influence of copper and temperature on the incidence of mortality due to dragonfly attack, we used the number of fatal attacks per tank per trial as a response and included all treatments that included a predator as input data (i.e., half of the treatments in the experiment). In this analysis, we modeled copper and temperature as main effects and used dragonfly size and developmental stage as fixed effect covariates. These models also used a count as a response and assumed a Poisson distribution.

Time to first attack

We tested how the factors in our experiment influenced the time from the beginning of the trial to the first attack made on a tadpole as a metric by which we could gauge relative dragonfly behavior across treatments. In particular, we were interested in how our test factors and covariates influenced dragonfly willingness to initiate an attack. We evaluated the effect of our main factors (copper and temperature) using dragonfly size and tadpole developmental stage as covariates. We used only data from the subset of treatments in which an attack occurred (i.e., those that contained a predator that at some point during the trial attacked a tadpole). These models assumed the response variable was distributed normally.

Tadpole activity

The behavioral analysis enabled us to use data from all of our treatments to test the main effects of copper, temperature, and predator presence (and covariates of dragonfly size and tadpole developmental stage) on tadpole behavior. We first modeled tadpole activity levels, measured as the proportion of tadpoles in each tank moving versus still at evenly spaced 1‐minute intervals for 120 observations per tank in each 2‐hour trial period. These models utilized all experimental data to allow us to compare the presence versus absence of predators on these behavioral responses. After reviewing the video recordings, we realized through over 12,400 point observations that a fatal attack in the treatment tank was a strong determinant for the behavior of the remaining tadpoles. Therefore, we also ran a separate series of models on half the data set in which predators were included in the tanks to enable us to test the occurrence of a fatal attack in the tank as a covariate. In these follow‐up models, we used copper, temperature, and the occurrence of a fatal attack in a tank, developmental stage, and dragonfly size to predict tadpole activity levels. Copper, temperature, fatal attack (if one occurred in the tank as a binomial predictor), tadpole developmental stage, and dragonfly size were all treated as fixed effects. This analysis used logistic regression models.

Tadpole position

We designed the experiment to test the main effects on tadpole vertical position, because prior studies have shown tadpoles threatened with predation to have a diving response (Peacor 2006, Fraker 2008) and tadpoles exposed to copper to have a surface preference (Reeves et al. 2011). We first used the entire data set to evaluate temperature, copper, and predator presence on tadpole vertical position, measured as the proportion of the 5 tadpoles in each tank in the top half versus the bottom half of the tank. Tadpole developmental stage and dragonfly size were considered as fixed effect covariates. As with tadpole activity, we discovered during observation of the video recordings that the occurrence of a fatal attack in the tank was a strong determinant of tadpole vertical position. We therefore performed a second analysis of the subset of our data in which a fatal attack was possible (all tanks containing a predator) to test the effect of a fatal attack on this response in the tadpoles that remained. These models were all logistic regressions due to the proportional nature of this response variable.

Dragonfly behavior

Finally, we wanted to understand how our main effects of copper and temperature influenced dragonfly behavior, which we measured as the proportion of the 120 point observations per two‐hour time period in each trial in which a dragonfly was moving (versus still). We used the same main effects and covariates as the attack models, and the input data included all treatments that contained a predator. The dragonfly behavior models also used logistic regression assuming an underlying binomial distribution to the data.

Results

We found our main effects of copper and temperature to influence predation dynamics and the behavior of each species individually. The results of each of our six analyses are described below. Model parameter estimates and test statistics for all analyses are given in Tables 1–6 and results are shown in Figs. 1–6. All AIC values are given in Table 7.

Number of fatal and nonfatal attacks in the experiment (13 trials, 520 tadpoles, 52 dragonflies). Copper increased the number of attacks (p < 0.01), warmer temperature increased the number of attacks (p < 0.01), and larger dragonfly size increased number of attacks (p < 0.01). Model results shown in Table 1.

The effects of copper, temperature, and fatal attack on dragonfly movement. There were five tadpoles and one dragonfly per treatment and observations of movement were recorded every two minutes (n = 6,240). Y‐axis shows the percent of observations in which dragonflies were moving. Copper decreased dragonfly movement (p < 0.03), warmer temperatures decreased dragonfly movement (p < 0.001), and a fatal attack increased dragonfly movement (p < 0.001). Model results shown in Table 6.

Summary of model parameters used in the best‐fit model for dragonfly activity.

Results for all models tested.

Total attacks

Copper, temperature, and dragonfly size each independently altered the number of total attacks made on tadpoles in the best supported model (Table 7). There were more attacks in copper‐treated tanks than in untreated tanks and in elevated temperature tanks than in ambient temperature tanks (Table 1, Figs. 1 and 2). Larger dragonflies attacked more often, but an effect of tadpole developmental stage was not supported. No interaction between the main effects of copper and temperature was supported in this analysis.

Average number of attacks per treatment. Copper increased the number of attacks (p < 0.01), warmer temperature increased the number of attacks (p < 0.001), and larger dragonfly size increased number of attacks (p < 0.01). Model results shown in Table 1.

Fatal attacks

Although there were interesting trends regarding copper and higher temperatures increasing the numbers of fatal attacks made on tadpoles (Fig. 1), none of our predictor variables were significant in this analysis (Table 2), probably due to a low overall count of fatal attacks in the experiment. There were only 19 fatal attacks made in 17 experimental units during the entire experiment (13 trials and 52 total tanks that included predators). Nearly half of these fatal attacks (9 of 19) were made on tadpoles in the elevated temperature treatments that included copper, and both tanks in which more than one tadpole was eaten during a trial were assigned to the high temperature with copper treatments. Of the remaining fatal attacks, four were made in treatments at ambient temperature with copper, and three each were made in the two treatments that did not contain copper. Moreover, there were 6 tadpoles eaten by dragonflies during the overnight acclimation period, because they somehow made it through the grate that separated the tadpoles from their predators. Of these, half were in the elevated temperature treatment with copper. Despite these interesting trends, we did not have the statistical power to detect a significant effect, likely due to such a low overall frequency of fatal attacks during our study.

Summary of model parameters used in the best‐fit model for fatal attacks.

Time to first attack

Temperature was the only factor that significantly altered the time to first dragonfly attack on a tadpole (Table 3, Fig. 3). Dragonflies in high temperature treatments (22°C) attacked tadpoles more quickly than dragonflies in ambient temperature treatments (17°C). Despite having a slightly shorter mean time to attack in treatments including copper, this effect was not large enough to be significant (Table 3, Fig. 3).

Time to first dragonfly attack on a tadpole. Only tanks with an attack (n = 28) are represented. Elevated temperature decreases time to first attack (p = 0.02). Model results shown in Table 3. Y‐axis shows model estimates (lsmeans) for the average time to first attack in each treatment.

Summary of model parameters used in the best‐fit model for time to first attack.

Tadpole activity

Copper, temperature, predator presence, and the occurrence of a fatal attack all independently altered tadpole activity (Table 4, Fig. 4). Copper had the greatest effect on tadpole activity, significantly decreasing activity levels even at low and environmentally‐relevant concentrations of 1.85 μg Cu/L (Table 4, Fig. 4). Warmer temperature increased tadpole activity, and the presence of a predator reduced it (Table 4, Fig. 4). Across treatments that included predators, the occurrence of a fatal attack further reduced tadpole activity, which spanned a range from 47% of tadpoles moving in high temperature treatments with no copper and no fatal attack to just 19% of tadpoles moving in ambient temperature treatments with copper in which a fatal attack occurred (Table 4, Fig. 4). Neither covariate of dragonfly size nor tadpole developmental stage was supported in this analysis, nor was an interaction between the main effects of copper and temperature.

The effects of copper, temperature, and fatal attack on tadpole movement. There were five tadpoles and one dragonfly per treatment and observations of movement were recorded every two minutes (n = 12,480). Y‐axis shows model estimates (lsmeans) for the average proportion of tadpoles moving in each treatment. Copper decreased tadpole movement (p < 0.001), warmer temperatures increased tadpole movement (p < 0.001), presence of a predator decreased movement (p < 0.001), and a fatal attack decreased tadpole movement (p < 0.001). Model results shown in Table 4.

Summary of model parameters used in the best‐fit model for tadpole movement using (A) all data and (B) data from predator treatments, in which a fatal attack could occur.

Tadpole position

Copper, predator presence, and fatal attack occurrence all influenced the depth of tadpoles in the water (Table5, Fig. 5). Across all treatments, tadpoles spent the most time at the top of the tank in the presence of copper and absent a predator (Table 5, Fig. 5). When a predator was present in a tank, a fatal attack in the tank was the largest driver of diving behavior, causing the remaining tadpoles to spend more time at the tank bottom. This effect was slightly reduced in copper treatments, where tadpoles showed the diving response less strongly (Table 5, Fig. 5). Temperature was not related to tadpole vertical position, and neither covariate of dragonfly size nor tadpole developmental stage was supported in this analysis. No interaction between the main effects of copper and temperature was statistically supported.

The effects of copper, temperature, predator presence, and fatal attack on tadpole vertical position. There were five tadpoles and one dragonfly per treatment and observations of movement were recorded every two minutes (n = 12,480). Y‐axis shows model estimates (lsmeans) summarized across temperature treatments because temperature was not a significant predictor of vertical position in these models (p = 0.67). Copper increased tadpole time spent at the surface (p < 0.001), presence of a predator decreased time spent at the surface (p < 0.001), and a fatal attack decreased time spent at the surface (p < 0.001). Model results shown in Table 5.

Summary of model parameters used in the best‐fit model for total tadpole vertical position using (A) all data and (B) predator data.

Dragonfly activity

Temperature and copper both influenced dragonfly activity levels, so that both warmer temperatures and the presence of copper decreased dragonfly activity (Table 6, Fig. 6). The effect size for temperature was greater than that of copper. The most pronounced predictor of dragonfly activity level in this study, however, was a fatal attack on a tadpole. In the 17 experimental units where dragonflies attacked and killed tadpoles, dragonfly activity levels were much higher than in treatments where fatal attacks did not occur. In two of these cases (both in the elevated temperature with copper treatment), dragonflies continued to hunt for new prey even after the first successful capture, killing and eating two tadpoles during a trial.

Discussion

Copper exposure at elevated temperatures increased the average numbers of dragonfly attacks on tadpoles by a factor of nearly three compared to copper free, ambient temperature controls (Figs. 1 and 2, Table 1). Trends in fatal attacks—though not significant—followed those of total attacks, with nearly half of the fatal attacks occurring in the elevated temperature with copper treatments (Fig. 1, Table 2). Interestingly, tadpoles in the elevated temperature with copper treatments suffered the most attacks, despite dragonflies exhibiting the least activity, and consequently we would expect that dragonflies expended less energy to capture more tadpoles in these treatments (Fig. 6). Moreover, dragonflies in the elevated temperature with copper treatments attacked tadpoles more quickly, with the first attack (on average) occurring 4 minutes after the trial began compared to 34 minutes after in the ambient temperature copper free controls (Table 3, Fig. 3). Based on these results, we predict that warmer and more degraded water quality—conditions projected to occur under global change scenarios—will cause dragonfly predators to have greater capture success for L. sylvaticus tadpole prey. Although nonfatal attacks may not directly decrease tadpole populations, they can cause injuries and malformations that may also be detrimental to survival (Ballengée and Sessions 2009, Bowerman et al. 2010, Reeves et al. 2010). Our observations provide insight into this predator‐prey relationship under varying environmental conditions, particularly those projected to occur more frequently under global change scenarios (Parmesan 2006, Heathwaite 2010, Roy and Bickerton 2012).

The differential effects of copper and temperature on the behavior of both dragonflies and tadpoles provides evidence into the mechanisms behind the attack data. Copper significantly reduced both dragonfly and tadpole activity levels (Figs. 4 and 6), but the effect size with tadpoles was much larger than with dragonflies whose activity levels were more influenced by temperature (Tables 4 and 6). In prior research, Reeves et al. (2011) demonstrated that 5 μg Cu/L significantly reduced tadpole activity, but the exposure concentration in the current study (1.85 μg Cu/L) was much lower. Our results supported these prior findings that extremely low concentrations of copper can reduce tadpole activity.

Yet our study added a new test factor, elevated temperature, which increased tadpole activity levels (Fig. 4, Table 4). Increased tadpole activity in warmer water has been linked to increased predator‐prey interactions and may therefore be maladaptive to tadpoles faced with visual predators, such as dragonflies (Skelly 1994, Anderson et al. 2001). Consistent with these studies and our predictions, more activity in the elevated temperature treatments increased the number of attacks on tadpoles (Figs. 1 and 2, Tables 1 and 2). Interestingly, dragonfly movement decreased in warmer water (Fig. 6). We hypothesize that dragonflies were less active in warmer water because the more active tadpoles were easier to catch. In the elevated temperature treatments, these visual predators may have had to move significantly less to find and attack their more active prey. Moreover, food consumption, assimilation, and metabolism of insect predators all increase at higher temperatures, causing these predators to need to feed more often to meet their metabolic needs (Anderson et al. 2001). Temperature may have a greater effect on predator metabolism or prey activity (and therefore detectability) than on prey escape ability, all of which may translate into more prey captured at warmer temperatures (Anderson et al. 2001).

Our findings support the results of prior studies that show tadpoles reduce activity and dive in response to a predator (Peacor 2006, Relyea 2007, Fraker 2008). Both responses were measurably stronger, however, in tanks in which a fatal attack occurred (Tables 4 and 5, Figs. 4 and 5). Moore and Townsend (1998) found tadpoles that were more active and spent more time at the surface were more likely to be eaten by a predator, suggesting that activity reduction and a bottom‐preference is an adaptive response. If this is the case, then we found copper to induce a maladaptive behavioral response in tadpoles also challenged by predators.

Copper‐exposed tadpoles spent more time at the water surface, in tanks with and without predators (Fig. 5, Table 5). This surface preference of copper‐exposed tadpoles replicates a result of Reeves et al. (2011) that copper exposure caused tadpoles to reduce activity and spend more time at the water surface. Both sets of results support a hypothesis that sub‐lethal concentrations of copper may inhibit larval amphibian ability to execute appropriate predator avoidance behaviors.

Reeves et al. (2011) predicted, but failed to detect, a diving response to chemical predator cues. Our results might seem to contradict these findings because in this study predator presence and behavior had a stronger effect on tadpole vertical position than did copper (Table 5, Fig. 5). Nevertheless, our finding that tadpoles responded quantifiably more strongly to an actual predation event than to predator presence alone may lend insight to, rather than contradict the prior result (Reeves et al. 2011). Here, we provide evidence that wood frog tadpoles may differentiate threat risks and thus respond differently to different levels of threat. Tadpoles in this study exhibited a stronger behavioral response to an actual predation event than to mere predator presence (Table 5, Fig. 5). Similarly, tadpoles may show a stronger behavioral response to actual predator presence than to chemical predator cues, which could explain the lack of the predicted diving response in the Reeves et al. (2011) study. Recently, Preston and Forstner (in press) also demonstrated a differential response of Bufo nebulifer tadpoles to predation (alarm) cues versus (kairomone) cues of predator presence, and showed a modulation of tadpole behavioral response based on aggregation status (one versus many tadpoles). Aggregated tadpoles had a greater tolerance for depredation risk. Reeves et al. (2011) also used a higher concentration of copper, which may have enhanced its effect on vertical position relative to the effect of predation cues in their study.

Based on a field study concurrent to this experiment, the copper concentrations and temperature treatments we administered are likely to be highly ecologically relevant. In a concurrent field examination into the mechanisms of frog malformations on the Kenai Refuge, dissolved copper was measured in 426 individual water samples collected at 36 Kenai Peninsula wetlands (including those from which these eggs were collected) between 2010 and 2012 (doi:10.5061/dryad.sq72d). Copper was detected at 64% of these sites, but only in 16% of samples. The highest analytical limit of detection for copper in this dataset was 1.85 μg/L. In a sensitivity analysis, we substituted half of the highest analytical detection limit for copper and half of the dosing concentration for this study (0.93 μg/L) for all samples measured to be below the detection limit. We then compared all copper values to the hardness‐based water quality standard for aquatic life provided by the ADEC (2008). We found that copper exceeded the hardness‐based chronic water quality criteria for aquatic life in 28 of 36 sites (78%) with exceedences shown for 203 individual samples. Similarly, 26 of 36 sites showed at least one exceedence of the acute standards for copper with 178 individual samples over acute toxicity thresholds. Copper exceeded chronic toxicity thresholds in 13 sites where it had never been detected, because the substitution value of half the detection limit (0.93 μg/L) was high enough to exceed the hardness‐based water quality criteria at these sites, due to softness of site water. Therefore our target dose of 1.85 μg/L copper for this experiment has demonstrated biological effects of copper on the predation dynamics of tadpoles and dragonfly larvae at extremely low copper concentrations, essentially at the analytical detection limits. Water hardness in this experiment was basically immeasurable—both Ca and Mg were below the instrument's limit of detection (LOD) in the confirmation sample (Ca LOD = 94.81 μg/L and Mg LOD = 152.88 μg/L). Yet, similarly low hardness values are commonly seen in Kenai Peninsula study sites where this same data set of 426 water samples had hardness values ranging from below the LOD for both Ca and Mg to 140.87 mg/L, with a median value of 8.67 mg/L. The results of the present study confirm the mechanisms proposed in our prior laboratory experiment (Reeves et al. 2011), and support the low toxicity thresholds for the State of Alaska's hardness‐based water quality standards for copper using experimental data.

Conclusions

Natural stressors, like predators, can combine with anthropogenic stressors like contaminants to create a more lethal environment for amphibians (Relyea 2004, 2005, Relyea and Diecks 2008, Rohr et al. 2008). The multiple stressors in this study altered species interactions in complex ways, changing the behavior of both predator and prey, and culminating in greater attack risk for tadpoles exposed to copper and elevated temperatures. We found that concentrations of copper approaching the analytical detection limit and temperatures slightly above average altered both predator and prey behavior, but in different ways, culminating in higher attack frequencies and greater ease of prey capture for the predator. Copper and temperature were shown to have different effects on tadpoles and dragonfly larvae, creating a complex relationship between predator and prey. With the large number of amphibian populations in decline, it is important to study these predation dynamics in aquatic ecosystems to understand what role this may play in causing these declines. This study reveals how two independent stressors can combine to have detrimental effects on amphibian prey. Based on the results of this study, as temperatures increase and contaminants enter aquatic systems, predation risk for tadpoles from larval dragonflies could increase.

Acknowledgments

This work was funded by the U.S. Fish and Wildlife Service, Department of Environmental Quality (FFS#7N26 and DEC ID#201070001.3). The findings and conclusions in this article are those of the author(s) and do not necessarily represent the views of the U.S. Fish and Wildlife Service. We thank Dr. Roman Dial for discussions that improved this study and feedback on an earlier version of this manuscript.

Supplemental Material

Appendix

Additional Methodological Information

Animal husbandry

The four aquaria were outfitted with aerators and kept outside in a covered, secure area. Initial water temperatures were approximately 4°C in the morning and 8°C in the evening and were consistent across aquaria. Water temperatures at this point ranged from 5°C to 11°C throughout the day. Two days after egg masses were mixed, some egg mortality was observed in each aquarium, but was minimal. Through the period of egg and early tadpole development, half of the water was changed every few days and water temperature was checked. After three weeks the total amount of water in each aquarium was increased to 5 L. As the eggs hatched at the beginning of June tadpoles were selected haphazardly and moved into additional aquaria, also outfitted with aerators, to allow more space for growth. Temperatures were checked every few days to ensure conditions were consistent across tanks and roughly half of the water was removed and replenished every few days. Water temperatures through the first several weeks of June were measured to be 14°C in the outdoor tanks. Once inside, tank water was replenished every 3 days, using roughly 6 L of Alaska's Best Water. The water was also changed completely every 2 weeks to limit algal growth. In an effort to control for potential covariates, all tadpoles and dragonfly larvae were measured and assessed for injuries prior to the start of the experiment. Only uninjured tadpoles and larvae were subject to experiments. All trials were conducted when tadpoles were between the developmental Gosner stages 26–41 (Gosner 1960), from limb emergence until the beginning of metamorphosis (Reeves et al. 2011). Once they reached the proper developmental stage, five tadpoles were placed in one side of each 2.5‐gallon container for the duration of the experiment, divided in half by a grated partition. One larval predator was added to the other half of the container in the appropriate treatments. Both species acclimated to these surroundings for twenty‐four hours before the experiment began. After the acclimation period, the tadpoles and dragonfly were placed together in one half of the aquarium. Confirmation samples were taken for copper‐treated and control water.

Copper

The control sample measured less than the detection limit of 0.38 μg/L. The copper‐treated sample measured 1.02 μg/L Cu, with an analytical detection limit of 0.38 μg/L. Because this was lower than expected, we re‐ran the sample. On the second run, the sample measured 1.34 μg/L Cu. This close to the instrument detection limits, and error of 30% is common (B. Hagerdorn, personal communication), but our exposure concentration may have been lower than our target dosing concentration by approximately 0.5 μg/L.