Resident cutthroat trout in the central Coast Range of Oregon : logging effects, habitat associations, and sampling protocols

Graduation date: 1996Presentation date: 1996-01-23Populations of coastal cutthroat trout, Oncorhynchus clarki clarki, were sampled in 16 headwater streams from logged (20-30 and 40-60 years ago) and unlogged (stand age 125-150 years) basins. Basins logged 20-30 years ago supported the widest range of mean biomass of age 1+ or older cutthroat trout (g/m2) and the widest range in frequency of large woody debris (number of pieces/100 m) and pools (number/100 m), including the lowest and highest levels of these variables encountered in the study. Stream gradient and geology of the substrate were important stream characteristics that influenced the age structure of cutthroat trout in pools. Biomass of young-of-year (YOY) cutthroat trout (g/m2) was directly related to stream gradient and inversely related to the biomass (g/pool) of age 1+ or older cutthroat trout, especially in pools of sandstone streams. High use of pools without adult cutthroat trout by YOY suggests that when older cutthroat trout abandon pools, either by movement or mortality, important rearing\ud opportunities are created for YOY. I used the population size and variance estimators of Seber and LeCren (1967) and Zippin (1956) to develop tables for use in the field that list acceptable ranges of observed catch from two and three removal passes. The ranges of acceptable values were based on an estimate for coefficient of variation (CV) and, for data from three passes, the values were based on results of a chi-square test for the assumption of equal catchability. Because the full multi-termed version of Zippin's (1956) estimator of variance (i.e., with Stirling's second-order approximation) was found to be more conservative than the use of his large-sample estimator when population size was less than 200, 1 used the full multi-termed version to derive estimates of CV. Because the distributions of population\ud estimates derived from removal data are typically not normal, I discuss the use of two other confidence intervals that do not depend on the assumption of normality, those based on the log-normal distribution or a profile likelihood ratio