Experiment One
Pesticide type explained 43 percent of the variation in Bi.
glabrata biomass (Table S4), whereas most of the variation in Bu.
truncatus biomass (35 percent) was explained by pesticide identity
nested within class nested within type (Table S4). Likelihood ratio
tests determined that the above pesticide effects were significant for
both snail species (Table S5). Model selection favored using pesticide
type and pesticide nested within class within type for Bi.
glabrata and Bu. truncatus biomass, respectively, and the lowest
AICc values included a term for zero-inflation and treating water and
solvent as a single control group for both snail species (Tables S6-S7).
Model selection (Table S6) indicated significant positive fertilizer
effects (bottom-up) on Bi. glabrata biomass (Fig. 1a; Table 1).
In contrast, bottom-up effects of herbicides for Bi. glabratabiomass were generally not significant in the presence of predators
(Table 1; p > 0.05). Insecticides were associated
with positive top-down effects on Bi. glabrata biomass (Fig. 1a;
Table 1). Similar to Bi. glabrata , fertilizers had significant
positive bottom-up effects for Bu. truncatus biomass, after model
selection (Table S7), while none of the six herbicides were
significantly associated with Bu. truncatus biomass in the
presence of predators (Fig. 1b; Table 1). Similar to Bi.
glabrata , insecticides were associated with positive top-down effects
on Bu. truncatus biomass (Fig. 1b). Four insecticides led to
significant increases in average Bu. truncatus biomass relative
to the controls (chlorpyrifos: 55% increase, terbufos: 36%,
esfenvalerate: 28%, and permethrin: 61%), while one insecticide from
each insecticide class had no significant effects (malathion and
λ-cyhalothrin both changed average biomass by < 10%).
Structural equation modeling (SEM) was consistent with the above
relationships (Fig. 2). The final SEM was a good fit for Bu.
truncatus (CFI = 0.99, χ2 = 15.25, df =
26, p = 0.95; Table S8) and Bi. glabrata (CFI =
0.99, χ2 = 28.04, df = 26, p = 0.36;
Table S9). The SEM model for Bu. truncatus suggested positive and
indirect top-down effects of insecticides on snail biomass by lowering
24 hr crayfish survival, which lowered final crayfish abundance (Fig.
2). Fertilizer increased both phytoplankton and periphyton chlorophyll
a, the latter of which had positive but non-significant effects upon
snail eggs and biomass (Table S8). Instead, significant positive
bottom-up effects of fertilizer occurred via increasing submerged
vegetation, which increased both snail eggs and biomass (Fig. 2). The
sum of standardized coefficients of indirect pathways from fertilizer to
snail eggs or biomass was over an order of magnitude greater via
submerged vegetation than via periphyton chlorophyll a alone. Submerged
vegetation was reduced by crayfish via consumption, which also formed a
significant and negative indirect pathway from crayfish predators to
snail eggs or biomass that was independent of crayfish predation on
snails (Fig. 2). Thus, insecticide-induced reductions in crayfish were
positively associated with snails both by reducing predation and
increasing snail habitat. The significance and direction of pathways in
the SEM model for Bi. glabrata were identical to Bu.
truncatus except that both snail eggs and submerged vegetation pathways
were negative for Bi. glabrata biomass (Table S9). Bi.
glabrata biomass at the end of the experiment was also approximately an
order of magnitude greater than for Bu. truncatus (Table S10).