CHINOOK SALMON, ONCORHYNCHUS TSHAWYTSCHA, 

 SPAWNING ESCAPEMENT BASED ON MULTIPLE MARK-RECAPTURE 



OF CARCASSES 



Stephen D. Sykes and Louis W. Botsford 1 



ABSTRACT 



Mark-recapture data from a population of chinook salmon, Oncorhynchus tshawytscha, carcasses were 

 collected for escapement estimates in a northern California stream. Escapement was taken to be im- 

 migration into the population of carcasses. Results from three methods of estimating total immigration 

 into this population— Jolly-Seber, Manly and Parr, and Jolly-Seber with a modified data set— were com- 

 pared to a weir count. Sources of violations of modeling assumptions, age-dependent catchability, and 

 survival were identified, but the estimates appeared to be relatively insensitive to these. The effect of 

 lower sampling intensity, which exacerbates effects of age-dependent catchability, was evaluated through 

 simulation. The third method appears to be the best of the three because 1) it requires the least sampling 

 effort, 2) it is the most robust with respect to violations of the assumption of equal catchability, and 

 3) it enables reanalysis of previously collected data. Standard errors and 95% confidence intervals of 

 estimates obtained by the third method were computed by simulation. Since the distribution of estimates 

 is asymmetrical, these confidence limits are preferred over standard expressions. 



Pacific salmon fisheries are currently managed by 

 attempting to allow a specified number of fish to 

 escape the fishery, migrate upstream and spawn. 

 Proper management therefore requires accurate 

 estimates of this escapement. Since Pacific salmon 

 die immediately after spawning, escapement can be 

 estimated from the number of carcasses that accu- 

 mulate during a spawning season. The California 

 Department of Fish and Game (CDF&G) estimates 

 escapement of chinook salmon, Oncorhynchus 

 tshawytscha, each year using the methods of Schaef- 

 fer (Schaeffer 1951; Darroch 1961) and Peterson 

 (Seber 1982) to analyze mark-recapture data from 

 surveys of accumulated carcasses. Since the fish 

 enter the stream to spawn during the sampling 

 periods, the assumption of a closed population re- 

 quired by the Peterson estimate does not hold. The 

 Schaeffer method is designed to estimate numbers 

 from a stratified two sample experiment in which 

 fish are tagged at different locations (or different 

 times at one location as fish migrate upstream) and 

 are sampled at the same locations (or an upstream 

 point) at a later time. CDF&G carcass surveys, on 

 the other hand, involve sampling the same unstrati- 

 fied stretch of spawning stream several times. The 

 results described here are part of an attempt to 

 develop an accurate, efficient, and robust procedure 

 for estimating escapement from carcass data. A 



department of Wildlife and Fisheries Biology, University of 

 California, Davis, CA 95616. 



Manuscript accepted July 1985. 



FISHERY BULLETIN: VOL. 84, NO. 2, 1986. 



technique that allows not just estimates for current 

 and future years, but also could be used to analyze 

 mark-recapture data taken by CDF&G in past years 

 was desired. 



Parker (1968) and Stauffer (1970) used standard 

 Jolly-Seber methods to estimate spawning run sizes 

 from mark-recapture data obtained from carcass 

 counts. However, they did not examine departures 

 from modeling assumptions by collecting appropri- 

 ate data in the field or statistically testing assump- 

 tions. Also, an independent count of the population 

 size was unavailable, hence actual errors in their 

 estimates could not be computed. In addition, car- 

 casses were carefully replaced where they had been 

 found after sampling and tagging, hence captured 

 carcasses would have a high probability of being 

 recaptured. Thus, their results were probably biased 

 because of heterogeneous capture probabilities. 



To develop the estimation technique a mark-recap- 

 ture experiment was performed in the Bogus Creek 

 spawning area of the Upper Klamath River drainage 

 during the 1981 chinook salmon spawning run. As 

 a check on the estimates, a counting weir was placed 

 at the mouth of Bogus Creek. Salmon were counted 

 while they were in the weir trap, and were sub- 

 sequently released upstream. This mark-recapture 

 study differed from the usual mark-recapture 

 studies of fish and wildlife populations in that the 

 population was composed of carcasses (i.e., in- 

 dividuals enter the population by dying and leave 

 by predation and decay). Thus, the age of a carcass, 



261 



