A Power Analysis on the Monitoring of Bull Trout Stocks Using Redd Counts

Abstract

The bull trout Salvelinus confluentus is listed as a federally threatened species in the Columbia and Klamath river drainages. A priori establishment of levels of decline or increase in bull trout redd numbers that will be considered biologically significant and levels of statistical significance that will be used to identify changes in redd numbers is essential to the success of future recovery plans. A prospective statistical power analysis indicates that with standard significance levels and two-tailed testing procedures the yearly variation in redd numbers typifying many stocks of bull trout in Montana limits the power of detecting less than 50% of changes in population size per generation to less than 0.8 during the first 15 years of a monitoring program. The limitations of monitoring bull trout stocks with redd counts, coupled with the critical nature of identifying future population changes, justifies the need to (1) identify and reduce the level of measurement error involved in redd counts, (2) use levels of statistical significance that adequately balance the risks of committing type I and type II errors, (3) use one-tailed testing procedures for identifying population declines during the initial and other critical years of a monitoring program, and (4) explore the use of other methods of monitoring.

Spatial and temporal patterns in fish assemblages of upper coastal plain streams, Mississippi, USA

“…A fish monitoring program focused on population trends may be overwhelmed by the observed background variation, even if a long time series of observations prior to the action of interest were available, which is rarely the case (Grossman et al, 1990). Only extreme trends in population levels would be detectable given the high variability we observed (Maxell, 1999). The claim is often made that at least large changes in abundance and species composition will be detectable by monitoring programs with minimal sampling prior to the action of interest.…”

“…Understanding and quantifying temporal variation is valuable for (1) clarifying sources of assemblage regulation (e.g. biotic versus abiotic) across stream types and regions (Grossman et al, 1998), (2) identifying species or assemblages with high rates of immigration and emigration, indicating dependence on habitat connectivity (Gowan et al, 1994), (3) designing appropriate monitoring and research approaches (Maxell, 1999), and (4) correctly interpreting time-series data on fish assemblages (Schlosser, 1990). Comparison of spatial and temporal variation can guide decisions regarding the most efficient distribution of sampling effort (e.g.…”

Bull Trout Trends in Abundance and Probabilities of Persistence in Idaho

“…A benefit of this approach is that the slope of the regression line fitted to the log e transformed abundance data is equivalent to the intrinsic rate of change (r intrinsic ) for the population (Maxell 1999) and produces unbiased estimates of r intrinsic despite the potential presence of observation error within the data (Humbert et al 2009). Values of r intrinsic less than zero indicate negative population growth, whereas r intrinsic values greater than zero indicate positive population growth.…”

“…Values of r intrinsic less than zero indicate negative population growth, whereas r intrinsic values greater than zero indicate positive population growth. We used a significance level α of 0.10 to increase the probability of detecting trends (Peterman 1990;Maxell 1999). Previous work has suggested that an inflection point in Idaho Bull Trout abundance corresponds to about 1994 (High et al 2008;Copeland and Meyer 2011), so we estimated r intrinsic for three periods: pre-1994, post-1994, and the entire time period.…”

Migratory Patterns, Structure, Abundance, and Status of Bull Trout Populations from Subbasins in the Columbia Plateau, Annual Report 2001-2002.

“…Detecting changes in population size may not be possible using the most extensive sets of redd count data available (7-17 years) (Maxell 1999) and is unlikely for populations for which more limited data sets exist (Rieman and Myers 1997). Errors in redd identification not considered in these earlier analyses may further limit the utility of redd counts.…”