Abstract
Accurate estimation of the age and sex composition of Pacific salmon Oncorhynchus spp. is essential for understanding their population dynamics. Unfortunately, the data collected from spawning ground surveys are generally biased by the probability of recovery. Using 11 years of mark-recapture data for fall chinook salmon O. tshawytscha in the Salmon River, Oregon, I developed logistic models to examine the influences of stream flow and the sex and size of the fish on the probability of carcass recovery from spawning grounds. The results indicated no significant differences in recovery rates for males and females. The probability of recovery generally increased as fish size increased and as stream flow decreased. However, the relationships were nonlinear; the probability leveled off for very large fish and increased slightly at very high stream flow. Directly computed age and sex compositions from survey samples underestimated the ages of small fish and males while overestimating those of large fish and females. For example, in 2000, age-2 jacks were underestimated by 75% while age-6 fish were overestimated by 21%, and male chinook salmon were underestimated by 8% while females were overestimated by 12%.