Detecting the response of fish assemblages to stream restoration: Effects of different sampling designs

Increased trout production within limited stream reaches is a popular goal for restoration projects, yet investigators seldom monitor, assess, or publish the associated effects on fish assemblages. Fish community data from a total of 40 surveys at restored and reference reaches in three streams of the Catskill Mountains, New York, were analyzed a posteriori to determine how the ability to detect significant changes in biomass of brown trout Salmo trutta, all salmonids, or the entire fish community differs with effect size, number of streams assessed, process used to quantify the index response, and number of replicates collected before and after restoration. Analyses of statistical power (probability of detecting a meaningful difference or effect) and integrated power (average power over all possible alpha-values) were combined with before-after, control-impact analyses to assess the effectiveness of alternate sampling and analysis designs. In general, the more robust analyses indicated that biomass of brown trout and salmonid populations increased significantly in restored reaches but that the net increases (relative to the reference reach) were significant only at two of four restored reaches. Restoration alone could not account for the net increases in total biomass of fish communities. Power analyses generally showed that integrated power was greater than 0.95 when (1) biomass increases were larger than 5.0 g/m(2), (2) the total number of replicates ranged from 4 to 8, and (3) coefficients of variation (CVs) for responses were less than 40%. Integrated power was often greater than 0.95 for responses as low as 1.0 g/m(2) if the response CVs were less than 30%. Considering that brown trout, salmonid, and community biomass increased by 2.99 g/m(2) on average (SD = 1.17 g/m(2)) in the four restored reaches, use of two to three replicates both before and after restoration would have an integrated power of about 0.95 and would help detect significant changes in fish biomass under similar situations.