Supplementary Results

Description: these results present a re-analysis of the data using only wetlands at which green frogs (Lithobates clamitans) were dissected (n = 28 wetlands). Response variables include larval trematode richness, overall trematode abundance, and the abundance of specific trematodes known to cause pathology in amphibians (Ribeiroia and the echinostomes). No re-analysis of Ribeiroia presence is presented as this variable is unlikely to be influenced by the amphibian species were dissected (i.e., both American toads and leopard frogs often become infected with Ribeiroia).

Model Selection and Analysis

Using only data from ponds with green frogs (Lithobates clamitans) (n = 28 wetlands), the best-supported model describing overall trematode species richness within amphibian hosts was model 16, which included the amount of forested area and factors related to intermediate hosts (snail density, snail richness, tadpole density and frog density). Models 3 (forested area) and 18 (distance to Mississippi, forested area, snail density, snail richness, and frog density) also had support (Supplementary Table 1). The amount of forested area had consistently strong and positive effects on trematode richness with a cumulative Akaike weight approaching 1 (0.99) (Supplementary Table 2).

For overall trematode abundance (aggregated among species), the best-supported models were 19, 17 and the Global model, which were within 1 AICC unit and included definitive host factors (forested area and proximity to Mississippi River), intermediate host factors (snail density and richness, tadpole density, and frog density in models 17 and Global), and physicochemistry (Global model only) (Supplementary Table 1). The most supported predictors within these models were proximity to the Mississippi (+), forested area (+), tadpole density (+), snail density (+) and snail richness (-) (Supplementary Table 2).

Models to predict individual trematode groups yielded similar results. The abundance of R. ondatrae was best predicted by models 18, 15 and 2, which included definitive and intermediate host factors: proximity to the Mississippi River (18, 15 and 2), forested area (18 only), the density and richness of snails (18 and 15 only), the density of frogs (18 and 15) and the density of tadpoles (15 only) (Supplementary Table 1). The most strongly supported variable was proximity to the Mississippi River (+) (Supplementary Table 2). The highest-ranking model predicting echinostome abundance also came from the definitive and intermediate host suite of models (Model 19). Once again, definitive host factors such as forested area and proximity to the Mississippi River contributed positively to parasite abundance, whereas snail and tadpole density were the most important intermediate host variables (each with positive coefficients) (Supplementary Tables 3-4).

Supplementary Table 1. Highest ranking models used to predict each response variable. Presented here is the “best” model for each response (based on AICC values) and any models within 2 AICC units of the best model. Descriptive diagnostics include the Residual Sum of Squares (RSS, for logistic regression analyses the -2loglikelihood is presented), the number of parameters (k), the Akaike Information Criterion adjusted for small sample sizes (AICC), the difference between a given model’s AICC value and the lowest AICC value (AICC), and the Akaike weight of the model (wi). For complete descriptions of each model, see Table 2 in the main manuscript.

Response / Model / Category / RSS / k / AICC / ∆ AICC / wi
Trematode richness / 16 / DH+IH / 0.535 / 8 / -198.052 / 0 / 0.3377
3 / SV / 0.698 / 4 / -196.998 / 1.053 / 0.1994
18 / DH+IH / 0.548 / 8 / -196.707 / 1.344 / 0.1724
Trematode abundance / 19 / DH+IH / 5.390 / 8 / -68.689 / 0 / 0.3093
1 / Global / 4.142 / 11 / -68.517 / 0.175 / 0.2838
17 / DH+IH / 5.072 / 9 / -67.674 / 1.016 / 0.1861
Ribeiroia abundance / 18 / DH+IH / 1.86 / 8 / -136.11 / 0 / 0.2656
15 / DH+IH / 1.86 / 8 / -136.08 / 0.0312 / 0.2615
2 / SV / 2.37 / 4 / -135.61 / 0.4964 / 0.2073
10 / DH / 2.31 / 5 / -134.11 / 2.0007 / 0.0977
Echinostome abundance / 19 / DH+IH / 6.012 / 8 / -62.574 / 0 / 0.5436

Supplementary Table 2. Model-averaged coefficients, standard errors and 95% confidence intervals for the best supported independent predictors for each response variable. Boldened variables are those with high cumulative Akaike weights and 95% confidence intervals that do not include zero (considered to have strong support). Unboldened variables are those with high cumulative Akaike weights but 95% confidence intervals that include zero (considered to have moderate support).

Response variable / Variable / Model-averaged coefficient / Weighted unconditional SE / Lower 95% / Upper 95% / Cumulative Akaike weight
Trematode richness / Forested area / 0.695 / 0.1801 / 0.3440 / 1.0463 / 0.999
Snail density / -0.273 / 0.1221 / -0.511 / -0.034 / 0.631
Trematode abundance / Distance to Mississippi / -0.642 / 0.3632 / -1.349 / 0.066 / 0.863
Forested area / 1.158 / 0.6933 / -0.193 / 2.511 / 0.934
Snail Density / 0.709 / 0.4177 / -0.104 / 1.524 / 0.911
Snail Richness / -0.036 / 0.0871 / -0.206 / 0.133 / 0.910
Tadpole density / 0.972 / 0.4994 / -0.002 / 1.945 / 0.898
Ribeiroia abundance / Distance to Mississippi / -0.503 / 0.197 / -0.888 / -0.118 / 0.986
Snail density / -0.210 / 0.224 / -0.647 / 0.227 / 0.593
Frog density / -0.242 / 0.300 / -0.828 / 0.343 / 0.564
Snail Richness / 0.076 / 0.048 / -0.018 / 0.1703 / 0.589
Echinostome abundance / Distance to Mississippi / -0.6228 / 0.369 / -1.343 / 0.0969 / 0.814
Forested area / 0.6869 / 0.613 / -0.508 / 1.882 / 0.748
Snail density / 0.8345 / 0.407 / 0.040 / 1.629 / 0.927
Snail richness / -0.0409 / 0.0918 / -0.220 / 0.1383 / 0.893
Tadpole density / 1.1349 / 0.5436 / 0.0746 / 2.195 / 0.945