Ecological Archives M080-014-A1

Mari K. Reeves, Peter Jensen, Christine L. Dolph, Marcel Holyoak, and Kimberly A. Trust. 2010. Multiple stressors and the cause of amphibian abnormalities. Ecology 80:423–440.

Appendix A. Supplemental methodological information for contaminants sampling, data reduction, and the toxicity experiment.

Contaminants Sampling and Analysis

Two sediment samples were collected from each pond, one for organic contaminant analysis and one for inorganic contaminant analysis, using the methods of Csuros (1994). Samples were homogenates pooled from three random locations in a pond. At each location, we sampled the top 20–30 cm of sediment. Shallow site samples were collected with hand-held scoops – stainless steel for organics and plastic for inorganics. Deeper sites were sampled with an Eckman dredge. Organic samples were homogenized in stainless steel bowls. Inorganic samples were homogenized in Ziploc bags. Prior to sampling each site, equipment was decontaminated by washing with Alconox and water, rinsing with deionized water followed by hexane and then acetone. Inorganic samples were analyzed at the Trace Element Research Lab (TERL) in College Station, Texas. Organic samples were analyzed at the Geochemical and Environmental Research Group (GERG) in College Station, Texas. Sample results were compared to sediment toxicity thresholds presented in the National Oceanic and Atmospheric Administration, Screening Quick Reference Tables (Buchman 2008). Water samples were also collected from study ponds in 2004 and 2005 to sample for total inorganic elements using standard field collection protocols (Csuros 1994) and inductively coupled plasma/mass spectrometry at TERL. Sample results were compared to water quality criteria presented in the National Oceanic and Atmospheric Administration, Screening Quick Reference Tables (Buchman 2008). The contaminants we measured (metals and chlorinated organic pollutants) should remain relatively consistent through time, with the exception of some of the lighter molecular weight aromatic compounds.

Reducing the complexity of the contaminants data

After screening contaminants data for toxicants above at least one established toxicity threshold, we used principal components analyses (PCA) to reduce the dimensionality of contaminants data separately for organic and inorganic contaminants. These groups were kept separate because organic and inorganic compounds may have different environmental sources, different environmental fate and transport, and different modes of toxicity. If a contaminant was not detected at a site, half the detection limit for that compound was used as a substitute.

Inorganic contaminants exceeding at least one toxicity threshold in water were aluminum, barium, cadmium, copper, iron and manganese. Elements exceeding at least one threshold in sediment included arsenic, cadmium, copper, iron, manganese, nickel, and zinc. Cu was only detected in water from one site, and was therefore excluded from the PCA. All elements in water and sediment arsenic were log transformed to improve linearity prior to PCA. These elements were then subjected to PCA to result in the inorganic vectors.

Organic contaminants exceeding at least one toxicity threshold included phenanthrene (a polycyclic aromatic hydrocarbon or PAH), polychlorinated biphenyls (PCBs) and the following organochlorine pesticides: aldrin, mirex, heptachlor-epoxide, dichlorodiphenyltrichloroethane (p,p’-DDT and metabolites), lindane (BHC and metabolites), and chlordane (and metabolites – See Appendix B: Table B1 for metabolites detected). For the pesticides, parent compounds and metabolites were summed prior to PCA because they were correlated and because this made data interpretation more straightforward.

For the metals, PCA vector 1 explained 33% of the variance and was positively correlated (r≥0.5) with the following elements: Iron, Manganese, and Nickel in sediment and Aluminum, Barium, Iron, and Manganese, in water. This vector was also negatively correlated (r ≤ -0.5) with sediment Arsenic and Cadmium (for correlations, see Table A1). PCA vector 2 explained 25% of the variance and was positively correlated (r ≥ 0.5) with Copper, Iron, Nickel, and Zinc in sediment. The third vector explained an additional 15% of the variance, but was redundant with the first two vectors and was therefore not retained for analysis. These PCA vectors were then used to represent the metals with which they were correlated in the regression analysis.

The organic data required several manipulations before PCA. First, one site was excluded from the organic PCA (and from all statistical analyses) because the sample was taken during a forest fire and concentrations of organic contaminants at this site were approximately an order of magnitude higher than all other sites, probably due to mobilization of these compounds by the fire and deposition in ash (Site KNA60; See Appendix B; Table B3 for data). Second, several organic contaminants representing a parent compound (e.g., DDT) and its metabolites (e.g., DDD and DDE) were quantified and reported separately by the analytical lab. We summed these chemicals (parent compounds and metabolites of DDT, chlordane, and lindane) before the PCA because we believed they would have similar environmental fate, transport, and relatively similar toxicological effects. Additionally, this step eased data interpretation. After these manipulations, we included the following organic contaminants which were over at least one toxicity threshold in at least one study site in the PCA: the PAH phenanthrene, total PCBs, and the organochlorine pesticides aldrin, heptachlor-epoxide, mirex, lindane (and metabolites benzene hexachloride or BHC), chlordane (and metabolites), and DDT (and metabolites DDD and DDE). We performed PCA on these untransformed variables. This resulted in the four components presented in Table A2. The first component explained 38% of the variance and was positively correlated (r≥0.5) with the following compounds: total PCBs, aldrin, heptachlor-epoxide, mirex, and chlordane. The second vector explained an additional 25% of the variance and was positively correlated (r ≥ 0.5) with lindane and DDT and negatively correlated (r ≤ -0.5) with total PCBs (Table A2). The third vector explained an additional 12% of the variance, but was only correlated with heptachlor-epoxide, which was also correlated with the first vector, and was therefore not retained for analysis. We therefore retained the first two PCA vectors to represent organic contaminants in the statistical analysis.

Site Sediment and Water Exposure Experiment

Sediments were collected from six sites in late April with hand-held stainless-steel scoops or Eckman dredge. Sediment samples were composites of three locations in a pond, homogenized in stainless steel buckets. Sampling equipment was decontaminated between sites by washing with Alconox and water, rinsing with DI water, then rinsing with hexane, then acetone, to remove organic contaminants and prevent cross-contamination between sites. Sediments were sorted to remove predatory invertebrates. On 12 May 2006, we collected six amplecting pairs of wood frogs from two breeding sites at which abnormalities have consistently been found and allowed them to oviposit in glass bowls. After oviposition, adults were released at their breeding sites and extra eggs were returned to the egg mass cluster. We collected site water twice per week with certified chemically-clean, 5-gallon cubitainers (Hedwin Corporation, Baltimore, Maryland, USA). The same cubitainer was used to collect water from each site at each water change. Water was changed every 4–6 days to prevent tadpoles from fouling the water. Old water was drained with site-dedicated siphon hoses, taking care to not harm the tadpole or disturb the sediment. It was replaced with temperature-equilibrated site water collected either that day or the day before. Water changes began after eggs hatched. Once tadpoles were free-swimming (Gosner (1960) stage 20), they were fed NASCO frog brittle for tadpole Xenopus, ad libitum at each water change. Four blocks, water baths with 24 bowls each, controlled the temperature of experimental units which were kept indoors under full-spectrum lighting on a light cycle simulating field conditions. Temperature was also set to mimic surface temperatures recorded in the field.

FigA1
 
   FIG. A1. Map of Kenai Study Sites (●). Heavy black lines are paved roads. Lighter black lines are gravel roads. Dark gray shading is KNWR boundary. Light gray shading is designated wilderness.

 

TABLE A1. Correlations between Metals PCA vectors and elements that exceeded toxic thresholds in sediment and water.

TableA1

 

TABLE A2. Correlations between Organic PCA vectors and chemicals that exceeded toxic thresholds in sediment and water.

TableA2

 

TABLE A3. Table of pairwise correlations between predictor variables used to model skeletal and eye abnormalities in wood frogs.

TableA3

 

LITERATURE CITED

Buchman, M. F. 2008. NOAA Screening Quick Reference Tables. NOAA OR&R Report 08-1, Seattle, Washington. Office of Response and Restoration Division, National Oceanic and Atmospheric Administration, 34 pp.

Csuros, M. 1994. Environmental sampling and analysis for technicians. CRC Press. Boca Raton, Florida, USA.


[Back to M080-014]