General
We established pond communities with variations in biotic and abiotic conditions that mirrored those found in the field. This was accomplished by the use of four different treatments that combined harvesting and the shading of communities: “Shading/Harvesting”, “No Shading/Harvesting”, “No Shading/No Harvesting” and “Shading/No Harvesting”. The treatments were designed to produce subtle effects on the biotic and abiotic conditions in the pond communities.
In the present study, we focused on genera from the family Daphniidae with different sensitivities to esfenvalerate (sensitive and insensitive D.). Long-term effects of three concentrations of esfenvalerate on populations of sensitive and insensitive D. were investigated for a period of 59 days after contamination. Changes in the structure and sensitivity of the whole communities are presented in the publication by Stampfli et al. (2011), in which only the treatments “No Shading/Harvesting”, “No Shading/No Harvesting” and “Shading/No Harvesting” were considered, as they represent a gradient of food availability and competition strength.
Microcosms: artificial pond systems
Ninety-six outdoor microcosms were installed at the Helmholtz Centre for Environmental Research in Leipzig, Germany (51°21′13 N, 12°25′55 E). For every concentration and treatment of shading and harvesting, six replicate microcosms were established (n = 24 per level of concentration). Each microcosm had a volume of 80 L and was filled with 60 L of water (tap water seeded with 1 L of natural pond water). The microcosms were maintained at this volume over the course of the experiment. Communities of freshwater zooplankton and sediment were collected from five different natural ponds within a radius of 15 km from the institute and established in the microcosms at the end of May and beginning of June 2008. The natural pond sediment was mixed at a ratio of 1:1 with sand and distributed on the bottom of each tank to a thickness of approximately 1 cm. Furthermore, approximately 10 g of shredded leaves (Populus spp.) were added to the microcosms. The collected organisms were distributed equally among all microcosms.
Awnings were positioned close to each pond at an angle of 45° so that the microcosms were shaded at around noon each day (12–4 p.m.). All microcosms were shaded for 4 weeks until 4 days before contamination to enable comparable communities to develop in all ponds. In microcosms subjected to harvesting, biotic interaction was reduced by removing 30% of the entire pond community each week using a net (10 × 12 cm, 250-μm mesh size). Organisms were harvested from 2 weeks before contamination and continued until the end of the experiment in September 2008. The harvesting was started 10 days before the removal of the awning for the “No Shading” treatments because we assumed that more time would be required for the invertebrates to adapt to the reduction in biotic interaction than for algal growth to adapt to the increase in light.
Pesticide exposure
Esfenvalerate, (αS)-α-cyano-3-phenoxybenzyl (2S)-2-(4-chlorophenyl)-3-methylbutyrate, is a synthetic pyrethroid that is widely used in agriculture and is highly toxic to aquatic insects and crustaceans. We used the commercial formula Sumicidin Alpha EC (BASF, Limburgerhof, Germany), which is an emulsifiable concentrate that contains 50 g/L of the active ingredient, esfenvalerate. On 4 July 2008, the microcosms were contaminated with three different concentrations (0.03, 0.3, and 3 μg/L) of the pesticide. The concentration of esfenvalerate decreased rapidly during the first hours in all setups. In addition, no significant differences in exposure among the different conditions of shading or harvesting were detected (for details, see Stampfli et al. 2011).
Biological sampling and environmental parameters
To determine species distributions and abundances, pelagic biological samples were collected and identified over the experimental period at the following time points: 13 and 5 days before contamination (mean: 9 days), and 4, 11, 16, 44, and 59 days after contamination. The samples were collected with a sampling tube (PVC, length = 31.7 cm, radius = 3.55 cm). The lid of the sampling tube was placed first in the centre of each pond on top of the sediment. Before the tube was fitted onto the lid, the water was stirred gently in order to obtain a homogeneous distribution of organisms in the pond. Afterwards, the water from the tube (which contained 1.7% by volume of the water from the pond), including any organisms, was passed through a sieve (180 μm mesh size). The organisms obtained in this manner were preserved in 70% ethanol, identified to the level of genus (Cladocera, Chaoboridae, Culicidae, Baetidae), order (Odonata, Copepoda) or class (Ostracoda, Arachnida) and counted under a microscope. The taxonomic groups that were relatively common in the pond communities are listed in Table 1.
Table 1 Abundances of main invertebrate taxa in the communities without pesticide exposure. The untransformed abundances are displayed with the mean and standard deviation from 9 days before until 59 days after contamination
Water temperature was recorded continuously with Handylog DK501-PL data loggers (Driessen & Kern, Bad Bramstedt, Germany). Differences in UV A + UV B radiation among the treatments were measured over the course of a sunny and a cloudy day in July with a UV meter (UV–VIS radiometer RM-21, Dr. Gröbel UV-Elektronik GmbH, Ettlingen, Germany). The presence of the awning reduced the radiation at the surface of the microcosms (average daily reduction due to the awning: 76% on both a sunny and a cloudy day). Water temperature also differed between the shaded and unshaded microcosms from the time at which the awning was removed until the last sampling point (minimum daily difference = −0.6°C, maximum daily difference = −3.3°C).
To monitor water quality in the different treatments, additional parameters were measured on a weekly basis for a subsample of 32 microcosms over the entire observation period. The additional parameters included the concentration of oxygen (WTW Multi 340i meter; WTW Instruments, Weilheim, Germany), pH (HI-98127; Hanna Instruments, Woonsocket, USA), electrical conductivity (HI-98312; Hanna Instruments, Woonsocket, USA), and the concentration of chlorophyll a as a measure of algal density (relative fluorescence units—RFU; GEMINI XPS Fluorescence Microplate Reader; Molecular Devices, Sunnyvale, USA). No differences in chlorophyll a concentrations were observed between shaded and unshaded ponds. However, in unshaded ponds oxygen concentration and pH were significantly higher (mean +23.8% and +3.5%, respectively) and electrical conductivity decreased (mean −6.8%). On the basis of these measurements of physicochemical parameters, we assume that shading has an indirect effect on algal growth (Anderson et al. 1994; Falkowski and Raven 2007).
Acute toxicity testing of esfenvalerate
Acute toxicity tests were performed to generate most comparable information on toxicological sensitivity of the Daphniidae present in the microcosms. The following species were tested: Daphnia longispina, Daphnia pulex, Ceriodaphnia reticulata and Simocephalus vetulus. The detected LC50 (96 h) values for the investigated species were similar to those previously published (Beketov 2004; Lozano et al. 1992; Werner et al. 2002). Not enough individuals of Scapholeberis sp. could be found for a toxicity test. For this reason we used the only existing literature value of LC50 (96 h) = 0.84 μg/L for esfenvalerate (Noskov 2011) to classify the genera.
Individuals of D. longispina, D. pulex, C. reticulata, and S. vetulus were collected in permanent and temporary ponds from the floodplains of the River Elbe, near Rosslau, Germany (51°53′06 N, 12°15′55 E), in June 2009. The organisms from the field were adapted to laboratory conditions in natural pond water under a constant air temperature of 20°C for 24 h before contamination with esfenvalerate. The pond water was passed through filter paper (mesh size: 1–2.5 nm) before the organisms were added for the toxicity tests. The electrical conductivity (EC) and pH of the used pond water were measured (HI-98312 and HI-98127; Hanna Instruments, Woonsocket, USA) and are provided in Table 2.
Table 2 LC50 values after 96 h with confidence intervals (CI) for the tested species and physicochemical parameters of the medium used
For the acute toxicity tests with esfenvalerate, we applied the following concentrations: 0, 0.003, 0.01, 0.03, 0.1, 0.3, 1, and 3 μg/L. Ten replicates per control and per concentration of esfenvalerate were used. Individuals were each kept in a volume of 50 mL of medium (pond water, described above) and monitored every 24 h until 96 h after contamination. After 24 h of exposure, the medium for all test samples and controls was changed to fresh uncontaminated medium. The LC50 after 96 h was calculated using the Trimmed Spearman–Karber method (Trimmed Spearman–Karber program, version 1.5, Hamilton et al. 1977).
Statistical analysis
The group of insensitive D. was generated by adding up the count data for all single genera in the family Daphniidae that were classified as insensitive taxa. Counted individuals and group data were fourth-root transformed, as suggested for skewed abundance data (Quinn and Keough 2002). Abundances of sensitive and insensitive D. were pooled for all treatments. Differences in mean abundance (n = 24 per concentration and control) at the various time points among the different concentrations of toxicant and the control were investigated with analysis of variance (ANOVA). The ANOVA was followed by pairwise t-tests for multiple comparisons and adjusted if the variances of the groups were not homogeneous. In the case of non-normally distributed samples, the Kruskal–Wallis test for nonparametric data was applied, followed by a nonparametric multiple-comparison test (R-package pgirmess, function kruskalmc; Siegel and Castellan 1988).
The influence of pesticide-related survival, 2 weeks after contamination and treatment of shading and harvesting, on the abundances of sensitive D. at the end of the experiment (6 and 8 weeks after contamination) was investigated with an analysis of covariance (ANCOVA). The pesticide-related survival was calculated as the ratio of the mean abundance from the samplings after contamination (11 and 16 days after contamination) to the mean abundance before contamination (−9 days) for each microcosm. Treatment was used as a categorical variable and pesticide survival of sensitive D. as a continuous variable. The models were simplified and validated in accordance with the work of Crawley (2007), by stepwise removal of nonsignificant terms until the minimal adequate model was reached.
Relations between abundances of sensitive and insensitive D. were tested for significance based on Pearson’s product-moment correlation for normally distributed data (correlation coefficient indicated with r) or Spearman’s rank correlation (correlation coefficient indicated with rho). Outliers were identified by checking correlations for noteworthy data points in fitted linear regression lines and applied model validation according to Crawley (2007).
We conducted a Principal Component Analysis (PCA) to assess correlations between sensitive D., insensitive D. and other taxonomic groups at pesticide concentrations with partial mortalities (0.03 and 0.3 μg/L). The selection of the linear multivariate method was based on the outcome of a preliminary Detrended Correspondence Analysis (DCA) following Leps and Smilauer (2003). The PCA was conducted and interpreted using correlation biplot scaling with centred and transformed species data (Zuur et al. 2007; Leps and Smilauer 2003). Species data were subjected to square-root transformation for reasons of most possible conformity with the previous univariate analyses. The concentration of the pesticide was log(x + 1)-transformed and added by passive ordination.
For the predicted long-term concentration–response curves we chose three abundances of insensitive D. 6 weeks after contamination, representing different percentiles of the observed abundances (“low” = 10th percentile, “medium” = 50th percentile, “high” = 90th percentile). The abundances of sensitive D. for three concentration–response curves were predicted, one for each scenario of abundance of insensitive D. The predictions on the abundance of sensitive D. at control and every concentration (displayed in % to control) were based on the regression lines that were fitted for relations between abundances of insensitive and sensitive D., 6 weeks after contamination.
Multivariate analyses were conducted using the program CANOCO 4.5 for Windows (Wageningen, Netherlands) in accordance with previous work and guides (ter Braak and Smilauer 2002; Leps and Smilauer 2003). The remaining statistical analyses and graphs were generated with R, version 2.11.1 (R Foundation for Statistical Computing, 2010).