Provided the estimates a, a, and E' are un- 

 biased, double inequality (4) gives an unbiased 

 estimate of the total mortality fraction from the 

 time of spawning to the tune of sampling. 

 Absence from the spawning bed of dead eggs 

 and larvae does not introduce bias to estimates 

 of Mi as it does to estimates of M T based on ratios 

 of dead to total eggs and larvae present. 



Potential egg deposition (E) is calculated by 

 multiplying the estimated number of female 

 salmon spawning within each area by average 

 fecundity. The value E' is obtained by multi- 

 plying E by a factor correcting for the fraction 

 of eggs and larvae present within the streambed 

 actually collected. In the present study, the 

 relationship 



(8) 



E '=ro E 



is used. Methods used to obtain the correction 

 9 



term. 



10 



are described elsewhere (McNeil' 



1962 a). 



More recent studies provide evidence that 



estimates of M, obtained with a hydraulic sampler 



are fairly representative of the true total mortality 



fraction. I used a hydraulic sampler to estimate 



total mortality of 1961 and 196.3 brood year 



preemergent pink and chum salmon fry in Sash in 



Creek, where total fresh-water mortality also 



was calculated from weir counts of adults entering 



and fry leaving the stream. In Sashin Creek, 



93 

 it was assumed that E' = jt-^ E for purposes of 



setting confidence limits to M t . Based on samples 

 of preemergent fry obtained with a hydraulic 



Table 2. — Total mortality of Sashin Creek pink and chum 

 salmon estimated by sampling preemergent fry with hy- 

 draulic sampler and by counting migrating fry at weir 



1 Mortality of the 1962 brood year was not estimated by sampling pre- 

 emergent fry because populations were very small. Only 4 pink and 42 

 chum salmon females entered Sashin Creek to spawn in 1962. 



sampler, confidence limits of M, bracketed the 

 total mortality fraction calculated for each species 

 and brood year from weir counts. In each instance, 

 mean estimated mortality from sampling pre- 

 emergent fry and from counting migrant fry 

 differed by less than 2 percent. The results are 

 summarized in table 2. In vSashin Creek, po- 

 tential egg deposition was determined by counting 

 adults entering the stream. In Harris River 

 and Indian and Twelvemile Creeks, the methods 

 of estimating potential egg deposition were not 

 as precise, and it is doubtful if estimates of M t 

 were completely unbiased. 



Presence or Absence of Eggs and Larvae 



Estimates of the population parameters M T and 

 M t sometimes fail to differ significantly when other 

 evidence suggests high mortality. A chi-square 

 test helped demonstrate significant mortality in 

 certain of these cases and often proved to be a more 

 sensitive test for detecting time of mortality than 

 the mortality estimates M t and M T - 



The chi-square test is based on the premise that 

 the proportion of points within a spawning bed 

 occupied by live eggs and larvae varies with total 

 mortality. If no change in mortality occurs, the 

 following conditions will be satisfied: (1) There 

 will be no decrease in the expected fraction of 

 points populated by eggs or larvae (live plus dead) ; 

 (2) there will be no decrease in the expected frac- 

 tion of points populated by live eggs and larvae ; 

 and (3) there will be no increase in the expected 

 fraction of points populated by dead eggs or larvae. 



In this study, each point sampled was classified 

 according to the number of eggs and larvae present, 

 with points containing less than 35 eggs and larvae 

 per m. 2 (k ) being classified together. The classes 

 used were (1) less than 35 live plus dead eggs and 

 larvae per m. 2 , (2) less than 35 live eggs and larvae 

 per m. 2 , and (3) less than 35 dead eggs and larvae 

 per m. L . 



Principal purpose of the classification scheme 

 was to classify jointly all points containing few 

 eggs and larvae and those containing none. The 

 selection of less than 35 per m. 2 for joint classifi- 

 cation was arbitrary, however. 



I tested each class independently with chi- 

 square (see Snedecor, 1956), and set confidence 

 limits to the number of points estimated to con- 

 tain fewer than 35 eggs and larvae per m. 2 from the 

 normal approximation of the binomial distribution. 



SPAWNING BED ENVIRONMENT OF PINK AND CHUM SALMON 



505 



