Experimental Design (Methods Section)

During summers 1999 and 2000, we placed 110-L plastic garbage bins near the headwaters of 10 naturally fishless tributaries of the East River that drain the steep eastern slope of Gothic Mountain at the Rocky Mountain Biological Laboratory in Western Colorado. We fashioned these bins with intake and outflow hoses so that fishless water from upstream was gravity-fed through them and then back into the stream at a rate of 2.5-3.0 L/min. We placed two brook trout (mean fork length +/- 1 SE = 155 +/- 6 mm in 1999; 169 +/- 4 mm, mean mass +/- 1 SE = 56 +/- 4 g in 2000) in the bins flowing into five randomly allocated streams, and the outflow of fishless stream water drained into the other five streams serving as controls. We fed these brook trout a mixed diet of stream invertebrates (including Baetis) every other day throughout the experiments. During summer 1999, the experiment ran from 30 July to 9 September to measure the effect of fish chemical cues on the size at emergence of [Baetis species B, which has a short larval period in the summer] (Peckarsky et al. 2001). In 2000, we ran the experiment from 29 June to 22 August to study the effect of fish chemicals on the size at emergence of [Baetis bicaudatus, which has a longer larval stage that overwinters as tiny larvae and develops to metamorphosis early in the following summer].

We delineated relatively low-gradient study sections of each stream from the outflow of the bins to 30 m downstream, which comprised about a third of the total length of each stream. In both experiments we collected mature Baetis larvae (with black wingpads: BWP) at weekly intervals, and measured their head capsule widths to the nearest 0.1 mm at 253 on a Wild dissecting microscope (Leitz Microsystems Incorporated, Bannockburn, Illinois) equipped with an ocular micrometer. We converted these linear measures to dry masses using regression equations for males and females (see Peckarsky et al. 2001:774), and analyzed treatment effects using dry masses. Mayflies with BWPs are within 24 h of emerging, have ceased feeding, growing, and have completed production of eggs (Peckarsky et al. 2001). Since adults do not feed, the size of BWP females is a good predictor of their fecundity (McPeek and Peckarsky 1998). We also measured sizes of BWP males, although large males do not have a mating advantage over smaller males (B. L. Peckarsky et al., personal observations). [Later published as Peckarsky, B. L., A. R. McIntosh, C. C. Caudill and J. Dahl. 2002. Stabilizing selection on male body size of high altitude populations of Baetis bicaudatus (Ephemeroptera: Baetidae). Behavioral Ecology and Sociobiology 51:530-537.]

We compared the size of mayflies that matured in fish-chemical treatment streams with those maturing in fishless control streams. In 1999, the [Baetis species B] began emerging 22 d after the start of the experiment, and we sampled individuals up to 43 d after the start of the experiment. Since the size of emerging [Baetis species B] remained constant over the emergence period in these and all streams previously studied (Peckarsky et al. 2001), we tested whether the mean size of BWP individuals differed among treatments. Due to low population densities (Peckarsky et al. 2001), and therefore small sample sizes of [Baetis species B], we used individual mayflies as replicates in a two-way ANOVA to determine the effects of treatment (fish odor or control) and sex (male or female) on the size at maturation.

In contrast to [Baetis species B], we have previously observed significant temporal changes in the size of emerging [Baetis bicaudatus] (Peckarsky et al. 2001). Furthermore, higher population densities of this [species] enabled us to collect sufficient numbers of mature individuals each week in 2000 to analyze effects of treatments on changes in Baetis size over time. In this analysis we used the mean size of individuals collected from each stream on each date as replicates, and time since the start of the experiment as a continuous variable. We tested for homogeneity of slopes of the relationship between time and mean size to determine whether the change in size of mature Baetis differed between streams that received trout chemicals and those that did not. We used maximum likelihood to compute all parameters including the variance components of a mixed model nested ANOVA (PROC MIXED; SAS 1989), with Baetis, sex, and fish odor treatment as fixed factors and stream nested within treatment as a random variable. Since mature individuals of both sexes were not obtained at every stream on every sampling date, we used a Satterthwaite correction for unbalanced designs to obtain the correct denominator degrees of freedom, which results in fractional df (see Table 2) (Searle 1987).

Smaller Baetis size at emergence and shorter development times in trout streams compared to fishless streams (Peckarsky et al. 2001) could be attributed to other differences between these two types of streams that we did not measure. Thus, we selected streams that were similar for both 1999 and 2000 experiments, and we measured attributes of the streams that could contribute to observed differences in size of mayflies at emergence. Hobo data loggers (Onset Computer Corporation, Pocasset, Massachusetts) were used to continuously monitor water temperatures in each stream, and we determined cumulative degree-days over 0 degrees C of control and fish-chemical treatment streams. We measured substrate particle sizes at 15 randomly chosen locations in each stream, and used a substrate index (Jowett et al. 1991) to summarize these data. We customized substrate categories for best resolution of variation among the streams as follows: bedrock, boulder = 20 cm, large cobble = 12-20 cm, small cobble = 5-12 cm, coarse gravel = 2-5 cm, fine gravel = 2 mm-2 cm, and sand = 2 cm. At each of these 15 locations, we also measured water depth, stream width, and current velocity, and estimated discharge at the beginning of each experiment. Finally, we measured conductivity in each stream using a YSI 30 meter (Yellow Springs Instruments, Yellow Springs, Ohio, USA).

To determine whether systematic habitat variation between streams allocated to different treatments could have confounded treatment effects, we conducted MANOVAs to test for differences between treatment and control streams in 1999 and 2000, including mean water temperature, discharge, substrate index, and conductivity. We also tested whether the treatment and control streams differed in algal biomass (food) and invertebrate densities (predators and competitors), since these variables have been shown experimentally to affect the development of mayflies (see citations in Peckarsky et al. [2001]).

Before starting the experiment, we estimated algal biomass (chlorophyll a) from 15 rocks in each stream, and used these data to demonstrate that resource levels did not vary systematically between streams allocated to the fish and fishless treatments (A. R. McIntosh and B. L. Peckarsky, unpublished manuscript). [Later published as McIntosh, A. R, B. L. Peckarsky, and B. W. Taylor. 2004. Predator-induced resource heterogeneity in a stream food web. Ecology 85:2279-2290.] We also took benthic invertebrate samples in all streams before the experiments, using modified Hess samplers in 1999 and electrobugging in 2000 (Taylor et al. 2001) to reduce disturbance of the substrates. The benthos of all streams was dominated by grazing mayflies, including Baetis and Cinygmula (Heptageniidae), and predatory invertebrates (primarily stoneflies), the densities of which were highly variable. We also used MANOVAs to determine whether densities of Baetis, other grazers, and predatory invertebrates differed among streams allocated to control and fish chemical treatments in 1999 or 2000.

Major Results

• Male and female Baetis of both species matured at smaller sizes in streams where fish chemicals were added.
• Fecundity was reduced of Baetis females of both species emerging from streams treated with fish chemicals.
• Natural variation in habitat did not contribute significantly to differences in observed Baetis size.
• Densities of predatory invertebrates were higher in control streams (not treated with fish chemicals), which could be a confounding effect.

Next step – Design Your Own Experiment (assignment).

Attribution

Many thanks to Marieke Perchik for contributing original artwork illustrating this study, to Emily Thorne for summarizing major results from this work, and to Dr. Barbara (Bobbi) Peckarsky for her review and suggestions.