Sources and Magnitude of Sampling Error in Redd Counts for Bull Trout

—Monitoring of salmonid populations often involves annual redd counts, but the validity of this method has seldom been evaluated. We conducted redd counts of bull trout Salvelinus confluentus in two streams in northern Idaho to address four issues: (1) relationships between adult escapements and redd counts; (2) interobserver variability in redd counts; (3) sources of interobserver variability; and (4) temporal and spatial variation in spawning activity. We found that estimated adult escapements and redd counts were strongly correlated on a logarithmic scale, but both sources of data probably contained large estimation or observation errors. In particular, redd counts varied significantly among observers in replicate counting trials. Observer counts ranged between 28% and 254% of the best estimates of actual redd numbers. Counting errors included both omissions and false identifications. Correlations between counting errors and redd and habitat characteristics were highly variable and provided limited insights into potential causes of sampling error. Finally, we found significant spatial and temporal variability in spawning activity, which should be considered in establishing index areas for redd counts and the timing of counts. Our results suggest substantial improvements are needed to make redd counts and unbiased estimates of adult escapement more useful for population monitoring. Bull trout Salvelinus confluentus are a conservation concern throughout most of their range in North America (Campbell 1998; U.S. Fish and Wildlife Service 1998, 1999). Much of what is known about the dynamics of bull trout populations comes from redd counts that have been used in a number of contexts, ranging from assessments of population trends and extinction risk (e.g., Rieman and McIntyre 1993; Rieman and Myers 1997) to analyses relating spatial variability in natural processes and human disturbance (e.g., Rieman and McIntyre 1996; Baxter et al. 1999). Redd counts offer several advantages for population monitoring. They are much less expensive than alternative methods used to inventory bull trout populations (e.g., underwater observation, tagging, trapping, weirs, and genetics). Thus, with limited resources it is possible to conduct redd counts for a greater number of populations over longer periods. Redd counts also may be less disruptive or invasive to bull trout populations than other methods. Finally, many time series of redd counts for bull trout now extend over 15 years, which exceeds available information on popula* Corresponding author: [email protected] Received January 7, 2000; accepted September 10, 2000 tion size from other sources (Rieman and McIntyre 1993, 1996; Rieman and Myers 1997). Most time series of redd counts are from populations of predominantly migratory bull trout, which are generally large (400–700 mm fork length; Rieman and McIntyre 1993); the females disturb a considerable amount of substrate during redd construction. These large excavations may be easily observed because spawning generally occurs from August to November (Rieman and McIntyre 1993), and stream discharges are usually low enough to allow good visibility of stream bottoms and redds. The apparent ease of identifying redds has led to a generally uncritical acceptance of redd counts for monitoring populations of migratory bull trout and many other species of sal-