Does regional diversity recover after disturbance? A field experiment in constructed ponds

Ethics statement

Our research was conducted at Washington University’s Tyson Research Center and approved by the Washington University in St. Louis Institutional Animal Studies Committee (protocol #20130058). All field research was conducted in accordance with the Missouri Department of Conservation guidelines under MDC Wildlife Collector Permit #15246. A licensed applicator (License #C18196) carried out the Rotenone application in accordance to the Missouri Department of Agriculture Pesticide Use Act.

Study site and experimental setup

This research was conducted in constructed ponds at Washington University’s Tyson Research Center (TRC) near St. Louis, MO, USA in 2012. In 2008, twelve ponds 6 m in diameter and 1 m in depth were constructed throughout the research center nested within similar terrestrial habitats of oak-hickory mixed forest. Ponds were separated by 500 m –800 m from each other and any other existing aquatic habitats. Ponds existed within a broader network of aquatic habitats, however, including temporary and permanent ponds, ephemeral and seasonal streams, and major rivers, the occupants of which provide the species pool for the experimental ponds in this study. At their start, ponds were stocked with a mixture of zooplankton and phytoplankton from more than 10 ponds in the region. Ponds were also stocked with gastropods and emergent and submergent macrophytes. Other macroinvertebrates (e.g., dragonflies, beetles, true bugs) were allowed to colonize naturally from the local species pool. Throughout the spring and summer of 2008, a large number of egg masses, larvae, and breeding adults of eight common species of anurans (Hyla versicolor, Acris crepitans, Pseudacris crucifer, P. triseriata, Rana sphenocephala, R. clamitans, R. sylvatica, and Anaxyurus americanus) were introduced from existing ponds at TRC to establish anuran populations. In 2011, at the conclusion of the initial research study conducted in these ponds, Ambystoma maculatum had naturally colonized 8 of the 12 ponds (Burgett, 2011). In order to ensure that there were no lingering effects of the previous study, we conducted pre-surveys of macroinvertebrates and zooplankton. The two treatments in this study (control and rotenone application) were stratified across past treatments to ensure that legacies of this past study would not bias our results. Initial pre-treatment sampling supported this, as we found no differences in aquatic community diversity or structure between ponds in the two treatment categories at the beginning of the study (see ‘Results’).

Rotenone treatment

Rotenone is commonly used as a pesticide and piscicide in North America. As a piscicide, it has been used for fish removal as a restoration strategy for small lakes, ponds, and other wetlands since the 1930s. Rotenone is most effective during the summer when water temperatures are above 21°C, which is also when zooplankton and macroinvertebrate richness and abundance are typically greatest. The half-life of rotenone in ponds is less than a day at 23°C (Gilderhus, Allen & Dawson, 1986; Gilderhus, Dawson & Allen, 1988), and it is largely considered the most benign fish control chemical agent widely used in management applications.

Rotenone was applied to six of the twelve ponds with the remaining six ponds left as untreated controls. A licensed pesticide applicator mixed a liquid formulation of Rotenone (Noxfish, 5% by volume active ingredient) solution to a 3 mg/L concentration, which was applied to each pond with a pesticide sprayer. This concentration is consistent with fish management strategies used in North America (Tate et al., 2003; Finlayson et al., 2010; Wynne & Masser, 2010). After application, the concentration of Rotenone that invertebrates were exposed to was approximately 0.15 mg/L. We sampled pond macroinvertebrates and zooplankton before Rotenone application on June 26th, 2012. The Rotenone treatment was applied on June 28, 2012. The first round of post-disturbance biodiversity sampling occurred on July 9, 2012 followed by a second round of biodiversity sampling 6 weeks post disturbance on August 13, 2012. We sampled macroinvertebrates and zooplankton again the following year on July 8, 2013 to monitor recovery.

Zooplankton sampling and data collection

We sampled zooplankton from ponds using a 2 L plastic pitcher. Eight 2 L samples of pond water were taken from the edge and center of the pond at varying depths and locations. We combined these samples and filtered them through an 80 µm zooplankton net to concentrate the sample to 50 mL. This sampling method is comparable to other studies in similar pond ecosystems (Steiner, 2004; Burgett & Chase, 2015). We preserved and dyed zooplankton samples immediately after collection using a standard Acid Lugol’s iodine solution.

We processed zooplankton samples after each field season. Each concentrated zooplankton sample was shaken and 10 mL was extracted and put into a petri dish. We used this 10 mL subsample to identify zooplankton species presence and abundance in a pond using a compound microscope at 20x–100x magnification. Species were identified to the lowest taxonomic level possible using keys from (Pennak, 1989; Thorp & Covich, 2010; Haney et al., 2013).

Macroinvertebrate sampling and data collection

We performed macroinvertebrate sampling using standard methods similar to those described in Chase et al. (2009). We placed a plastic cylinder stovepipe sampler measuring 36 cm in diameter and 1m tall in three randomly selected areas of the pond. We then used a 15 cm wide aquarium net to exhaustively sample macroinvertebrates from each stovepipe location. We sampled each stovepipe until 5 consecutive empty sweeps of the net. Every individual was collected and stored in one bin for all three stovepipes. In order to account for rare and fast swimming species, 20 additional net sweeps were conducted throughout the pond at varying depths and locations using a 30 cm-wide, 1 mm mesh D-net. These samples were collected in a separate container. We preserved all macroinvertebrates from these samples in 70% ethanol for later identification. We used a dissecting microscope for species identification to the lowest possible taxonomic group according to the keys of Needham, Westfall Jr & May (2000); Merritt, Cummins & Berg (2008); Thorp & Covich (2010).

Analyses

We combined zooplankton and macroinvertebrate data for each pond and each sampling period into species-by-pond matrices and conducted all analyses using these combined invertebrate (macroinvertebrate and zooplankton) biodiversity data. We quantified species richness as the number of species per pond (i.e., local richness) and using the Chao1 index (Chao, 1984) to account for any potential sampling bias. The Chao1 index estimates the number of species potentially missed during sampling based on the frequency of rare species in a sample. Chao1 values were calculated for each pond using the program Past (Hammer, Harper & Ryan, 2001). All analyses were run using both raw species richness and the Chao1 index.

We used T-tests in R (R Core Team, 2013) to verify that there were no differences in the number of species per pond between treatments prior to the addition of rotenone. In order to determine if rotenone addition impacted local richness both within and among sampling dates, we used linear mixed effects modeling (lme function in the nlme package in R (Pinheiro et al., 2013)). We chose a linear mixed effects modeling approach because local species richness values were normally distributed, and the mixed effects model accounted for the non-independence of the repeated measures of species richness in ponds over time. We selected the model with the lowest Akaike Information Criterion (AIC). The fitted model included fixed effects for differences in species richness between rotenone treatments, between sampling dates, and the interaction between rotenone treatment and sampling date, as well as the random effect of pond identity. The ANOVA results for the full model are provided in the (Table S1). When there were significant differences between treatments or among sampling dates, pair-wise multiple comparisons were used. Since the main results of this study are the same using Chao1 estimates of species richness or actual measured richness, only the results for the actual measured species richness are presented.

For analyses of pond species composition we calculated among-pond dissimilarity using Jaccard’s and Bray-Curtis dissimilarities for each treatment at each sampling date using the program PRIMER (Clarke & Gorley, 2006). We then used PERMANOVA to ask if pond invertebrate composition differed between control and rotenone ponds (Anderson, 2001; Anderson, Gorley & Clark, 2008; Chase, Burgett & Biro, 2010). We used a repeated measures PERMANOVA design with the fixed effects of rotenone treatment, sampling date, and the interaction between rotenone treatment and sampling date, and the random effect of pond identity nested within treatment. Post-hoc pair-wise PERMANOVA tests were used to identity if there were differences between control and rotenone ponds at each sampling date. Since the main effects of rotenone on species composition were the same using both Jaccard’s and Bray-Curtis dissimilarities, only the PERMANOVA results for Jaccard’s dissimilarity are presented. Finally, we visually assessed differences in species composition between treatments using nonmetric multidimensional scaling (NMDS) plots based on Jaccard’s dissimilarity in PRIMER (Clarke & Gorley, 2006).

We also aimed to test our prediction that species losses associated with the application with rotenone would be selective. Using R programming language (R Core Team, 2013), we applied a null model for random species extinction using the conceptual framework proposed by Smith, Lips & Chase (2009). This model develops a prediction for the expected number of extinctions under a null scenario of random extinction, given the magnitude of a series of local disturbances (in this case, rotenone). We compared actual species loss at the treatment level to this null prediction, which allowed us to determine if rotenone addition selectively removed species from local ponds resulting in a greater loss of regional species richness than would occur if species were randomly removed from local ponds. The number of species lost from each treatment pond between the pre sampling date and one week following rotenone addition was calculated and used to develop a random probability of extinction in the model. For each model run, the identity of the species that were removed (i.e., a local extinction) in each of the six-rotenone treatment ponds was chosen at random. The total regional (six pond) species richness was calculated after each run and saved. This model was executed 10,000 times, and mean expected regional richness and upper and lower 95% range values were calculated. We then compared these simulated values to our actual observed regional species richness after rotenone addition in order to obtain a p-value.

Local and regional species richness

Prior to the application of rotenone, there were no detectable differences in local or regional richness between the two treatment groups (Fig. 2, “Pre Rotenone”; local richness: NS, t = 0.41, df = 8, p = 0.55). On average, each individual pond had 17.4 invertebrate species (SD = 4.1), and the control treatment group of six ponds had a regional species richness of 49, while the rotenone treatment group had a regional richness of 50 species of invertebrates.

The addition of rotenone had a significant effect on local species richness (p < 0.008, F_1,10 = 11.20, Table S1; Fig. 2), but this effect varied across sampling dates (rotenone × sampling date interaction, p < 0.03, F_2,20 = 4.35, Table S1; Fig. 2). One week after the application of rotenone to the six treatment ponds, mean local species richness was 48% lower in rotenone treatment ponds than in control ponds (Figs. 2 and 3; t = 4.47, df = 10, p < 0.005). Rotenone treatment ponds had, on average, 9.5 species fewer than control ponds. After the rotenone treatment, regional species richness of treated ponds dropped by a large margin, almost 40%, having only 31 species across the six rotenone-treated ponds as compared to 50 species across the six control ponds (Fig. 2). These results are consistent with our hypothesis that a greater number of species would be lost from the regional than local scale following rotenone addition (Fig. 3).

Six weeks after the application of rotenone, mean species richness was still significantly lower in rotenone ponds than in control ponds (14.5 versus 18.8 species, respectively; t = 2.37, df = 10, p < 0.05, Fig. 2). However, paired t-tests revealed that local species richness of rotenone ponds was similar to species richness in both rotenone ponds prior to application of rotenone and rotenone ponds one week after application. Qualitatively, the recovery of species richness at the local scale was more moderate than at the regional scale, with increases of 4.5 and 13 species, respectively.

One year after the application of rotenone to half the ponds, mean invertebrate species richness was similar among control and treatment ponds and no longer statistically distinguishable (NS, t = 0.65, df = 10, p > 0.5, Fig. 2). Further, the regional species richness across the six rotenone ponds had increased to 54 species, exceeding the regional species richness of the control ponds at that date.

Species loss null model

The fact that the loss of regional richness was greater than the loss of average local richness suggests that rotenone selectively filters invertebrates from the affected ponds. This selectivity is further supported by comparison of actual regional species loss to our simulation of random species loss (Fig. 3). A random pattern of local extirpations would have resulted in a regional species richness of 40 species (mean: 39.61 95% CI [39.57–39.65]). However, significantly more species were lost from the six-rotenone treatment ponds than would be expected if local extirpations were nonselective (p < 0.0001).

Species composition

Selective extirpations among ponds would be expected to lead to differences in community composition between control and treatment ponds, which is what we observed (Fig. 4). Prior to the application of rotenone, ponds from both treatments were intermingled in ordination space and could not be distinguished (Fig. 4). We found a significant effect of sampling date (PERMANOVA, F_2,20 = 2.95 and P = 0.0001) and an interaction between rotenone and sampling date (PERMANOVA, F_2,20 = 1.56 and P = 0.0089) on invertebrate species composition. One week after rotenone application, control and rotenone ponds were significantly different in their invertebrate species composition as measured by Jaccard’s dissimilarity (Fig. 4; PERMANOVA, F_1,10 = 1.80 and P = 0.01, Table S2). Control and rotenone ponds continued to diverge in their species composition six weeks after rotenone application (Fig. 4; PERMANOVA, F_1,10 = 1.75 and P = 0.006, Table S2). One year after the rotenone treatment, invertebrate community composition was again indistinguishable between the two treatments, providing further evidence that re-colonization and recovery of species richness and pond community structure relative to control ponds had occurred.