Discrimination of Chinook Salmon, Coho Salmon, and Steelhead Redds and Evaluation of the Use of Redd Data for Estimating Escapement in Several Unregulated Streams in Northern California

—We developed and evaluated a stratified index redd area method to estimate Chinook salmon Oncorhynchus tshawytscha, coho salmon O. kisutch, and steelhead O. mykiss escapement in several coastal streams in northern California based on the assumption that redd size is related to the number of redds a female builds. Sources of error in redd counts were identified, including the use of logistic regression to classify redd species (necessary due to temporal overlap in the spawning of these species in coastal northern California). Redd area escapement estimates were compared with estimates from more conventional methods and releases above a counting structure. Observer efficiency in redd detection ranged from 0.64 (SE 5 0.10) to 0.75 (SE 5 0.14) and was significantly associated with streamflow and water visibility (analysis of variance [ANOVA]: F 5 41.8; P , 0.001). Logistic regression reduced uncertainty in redd identification. Redd area and date observed were significant in predicting coho salmon and steelhead redd species (Wald’s z 5 11.9 and 18.09, respectively; P , 0.001). Pot substrate and redd area were significant in classifying Chinook and coho salmon redds (Wald’s z 5 5.88 and 4.03; P 5 0.015 and 0.04, respectively). Stratified index redd area escapement estimates and estimates based on capture–recapture experiments, area-under-the-curve estimates, and known releases above the counting structure (coho salmon only) were not significantly different (ANOVA: F , 13.6; P . 0.06). Escapement estimates assuming one redd per female were only significantly different from other methods for steelhead (ANOVA: F 5 13.11; P 5 0.006). Redd counts were significantly correlated with escapement estimates (r . 0.82; P , 0.04). Reduction of counting errors and uncertainty in redd identification, biweekly surveys throughout the spawning period, and the use of redd areas in a stratified index sampling design produced precise, reliable, and cost-effective escapement estimates for Chinook salmon, coho salmon, and steelhead. Accurate estimates of escapement are essential for effective management and conservation of salmonids (Busby et al. 1996; McElhany et al. 2000). In northern California coastal Chinook salmon Oncorhynchus tshawytscha, coho salmon O. kisutch, and steelhead O. mykiss are listed as threatened species under the U.S. Endangered Species Act (U.S. Office of the Federal Register 1997, 1999, 2000). Estimates of abundance at the population level are likely to be an important, though not independent, part of delisting criteria. There is a * Corresponding author: [email protected] 1 Present address: Department of Mathematical Sciences, Clemson University, Clemson, South Carolina 29634-0975, USA. Received February 4, 2004; accepted June 8, 2004 Published online March 4, 2005 need for reliable, cost-effective, and precise techniques for monitoring salmonid escapement. While redd counts are commonly used to index adult escapement and assess population trends (Beland 1996; Rieman and Myers 1997; Isaak et al. 2003), their accuracy as a measure of abundance has rarely been evaluated (Dunham et al. 2001). As the product only of reproductive adults, redd counts provide an index of effective population size (Meffe 1986). Maxell (1999) suggests that the sources of counting errors involved in redd counts be identified and reduced before they will be useful for long-term monitoring. Dunham et al. (2001) suggest that redd counts are less intrusive and expensive than tagging, trapping, underwater observation, weirs, and genetics for inventorying bull trout Salvelinus confluentus populations, and that with limited resources more populations can 285 ESTIMATING SALMON AND STEELHEAD REDD ESCAPEMENT be inventoried over a longer period. However, they conclude that substantial improvements are needed to reduce counting errors before redd counts will be useful for population monitoring. The use of redd counts for population monitoring may be further complicated if females make more than one redd. Crisp and Carling (1989) found that female salmonids occasionally make more than one redd, and Reingold (1965) documents a steelhead making two redds in different locations within a stream. Salmonids may also make false or ‘‘test’’ redds, which are abandoned before eggs are deposited (Crisp and Carling 1989). The use of redd counts for population monitoring may be further complicated if there is uncertainty in redd species identification. Identification of Chinook and coho salmon and steelhead redds in coastal northern California streams is difficult because of overlap in spawning time and redd sizes. Chinook and coho spawn from late October through January, and steelhead spawn from December through March in coastal northern California (Weitkamp et al. 1995; Busby et al. 1996; Myers et al. 1998). Redd surface areas range from 0.84 to 15 m2 for Chinook salmon, from 0.80 to 8.4 m2 for coho salmon (Burner 1951), and from 2.4 to 11.2 m2 for steelhead (Orcutt et al. 1968). Thus, to use redd counts for population monitoring in coastal northern California it was necessary to develop a technique to distinguish redd species. To resolve some of the weaknesses listed above, we evaluated the amount of bias in estimates due to errors in redd species identification, detection of redds, and duration under variable survey conditions. We used data collected over 2 years in four rivers and three creeks to develop a logistic regression model based on physical redd characteristics and spawning time to distinguish between coho salmon and steelhead redds and tested it with data collected in the third year of the study. To distinguish between Chinook and coho salmon redds, a similar model was developed and evaluated with data collected in two rivers during 1 year. We evaluated the validity and some sources of bias involved with using redd counts and redd sizes to estimate escapement by estimating surveyor efficiency, the duration redds remain visible, and the influence of streamflow and water visibility on redd detection. To determine if redd-based estimates differed from conventional escapement approaches, we examined the relationship between these estimates and estimates based on capture– recapture experiments, area-under-the-curve (AUC) estimates, and counts at the Noyo River Egg Collecting Station (ECS) between 2000 and 2001 and between 2002 and 2003. To test if coho salmon and steelhead redd counts and redd-based escapement estimates are related to true abundance, we examined the relationship between these data collected over 4 years in one river (steelhead only) and 3 years in two rivers and three creeks. To determine if female salmonids make more than one redd we compared the number of redds observed to our AUC and capture–recapture estimates of the number of females. The two-fold purpose of this study was to determine if escapement estimates, based on redd counts or on the assumption that redd size is related to the number of redds a female salmonid makes, can be applied to all three species, and if they are more reliable, cost effective, and precise than conventional approaches.