414 



Fishery Bulletin 98(2) 



Characteristically, blue crabs have a patchy dis- 

 tribution, and the estimated CPUE from the dredge 

 survey is driven by relatively few large catches. To 

 estimate a mean catchability coefficient that is appli- 

 cable to the entire survey area, estimates of catcha- 

 bility from each removal experiment were weighted 

 by the abundance in the experimental area. An esti- 

 mator for the overall catchability coefficient to use 

 for calibrating CPUE in the dredge survey is 



' C 



where c, = the total number of crabs caught in the 

 (th experiment; 



g, = the corresponding estimated gear effici- 

 ency; and 



C = total number of crabs caught in the n 

 experiments. 



Because n is small within each year, the jackknife 

 estimate of average gear efficiency and its standard 

 error were used (Cochran, 1977; Efron and Gong, 

 1983). The jackknife estimator of standard error is 



:{[(/; -l)/«]^(0„,-0,,)-}' 



where 



^(n=X 



c,q, 





is the weighted mean catchability deleting the nth 

 experiment and 



e,, 



"-I^ 



is the jackknife estimate of q for the /; experiments. 



For model 2 we also estimated the weighted mean 

 and variance of the slopes from all n experiments. An 

 estimate of q was obtained after retransformation 

 with the method of Finney ( 1941 ); the standard error 

 was estimated by jackknifing (Tukey, 1958; Manly, 

 1997). 



In the annual winter dredge survey, a one-minute 

 tow is standard. For soft sediments, the dredge may 

 be saturated before the tow is completed. A rando- 

 mized block experiment was conducted during the 

 winter of 1992-1993 to investigate such gear-satu- 

 ration effects on CPUE. Double tows were made at 

 77 randomly selected stations in the Maryland part 

 of the bay. One tow of one-minute duration and one 

 tow of 30-seconds duration were taken in random 

 order at each station. 



During the winter of 1994-1995, the chicken wire 

 liner of the dredge was replaced with nylon mesh 

 because the latter proved to be easier to operate and 

 repair. To investigate any effect of the new liner on 

 the catchability coefficient estimate, we conducted 9 

 removal experiments using a dredge with a chicken 

 wire liner and 10 experiments using a dredge with a 

 nylon liner. 



Using the estimated catchability to calibrate CPUE 



If r is the true blue crab density (number of crabs 

 per m^) in Chesapeake Bay at the time of the winter 

 survey, and if we have an approximately unbiased 

 estimate of the overall catching efficiency q that is 

 uncorrelated with CPUE, an estimator for blue crab 

 density is then 



r = CPUE/q, 



where CPUE = the estimated mean number of crabs 

 caught per m- swept. 



Abaywide estimate of the population total ris 



Ar, 



where A = the area of blue crab habitat in Ches- 

 apeake Bay, which we estimated using 

 geographic information system (GIS) to 

 be approximately 11,000 km'-. 



Using Taylor series approximations, we estimated 

 the variance of r (Thompson, 1992, p 168) as 



var(r) : 



. _ (l-g)_ V" - - 

 var {y) + — -. — y -i- ^ var( q ) 



where y = the estimated CPUE. 



The relative precision of fis /? = Vvar(r)/ f. The vari- 

 ance of estimated catching efficiency thus adds to 

 the variance of absolute abundance estimates. 



If sampling fractions differ between strata, the 

 absolute abundance can be estimated separately for 

 each stratum. Here, q, CPUE, and hence f can be 

 estimated by stratum, by using the same approach as 

 above. The size of each stratum must also be known. 

 If the sampling in each stratum is independent, the 

 variance in the combined estimate of absolute abun- 

 dance is additive. This approach would also be appro- 



