Daniel J. Hornbach¹, Mark Hove¹, Jill Medland² and Randy Ferrin^2
1Dept. Biology, Macalester College, St. Paul MN.
²St. Croix National Scenic Riverway, St. Croix Falls WI

Sampling for endangered species is an extremely difficult task, especially in riverine systems where sampling involves SCUBA in areas of high current and low
visibility. There are also statistical issues surrounding sampling rare species. Managers must decide the level of population change and the degree of statistical
confidence that should be used to trigger management actions. If population density is low it takes a large sampling effort to detect small changes with great
confidence.

In 7 years of quantitative sampling at Interstate Park and Franconia, MN where Q. fragosa is found, Hornbach collected 4594 mussels from 30 species (mean
density=22 mussels/m²). Only 5 Q. fragosa were found. Given the rarity of Q. fragosa, it would require over 15,000 0.25 m² quadrat samples to detect a
significant change in the population density of Q. fragosa with any degree of certainty. This is an unacceptably high level of sampling activity.

On May 16, 2000 a team of managers and researchers met to discuss a sampling protocol for sampling Q. fragosa. Given the difficulties of sampling this rare
species quantitatively a two-step approach was suggested: 1. Sample the entire mussel community, quantitatively, to examine whether there are major shifts in
mussel density; 2. Qualitatively sample the Interstate Park area to determine whether the proportion of the community that Q. fragosa constitutes changes over
time. To focus the sampling effort we will limit sampling to mussels that contain pustules. Of these species, Q. fragosa constitutes 3.7% (95% confidence
range=2.5-5.0%).

We developed a resampling model that allowed us to examine the statistical power obtained with a range of sample sizes. Based on this model if Q. fragosa
constitutes as little as 2.5% of the mussels with pustules, we could detect a 20% decline in the proportion of the population constituted by Q. fragosa with 88%
confidence with a sample size of 2500. This represents a compromise between sampling effort and statistical power and confidence.