FISHERY BULLETIN VOL, 77, NO J 



ally consistent results. The method of Van Sickle 

 (19771 requires the most detailed information on 

 the size distribution and growth rates, and the 

 results are susceptible to the size groupings cho- 

 sen. Hancock's (1965) method requires less de- 

 tailed information on growth, as average annual 

 growth increments can be used, but still requires 

 an accurate description of the size-frequency dis- 

 tribution. Bhattacharya's (1967) method does not 

 require information on growth, although knowl- 

 edge of the annual growth increment does aid in 

 analysis of the data, but again requires an accu- 

 rate description of the size-frequency distribution. 



The instantaneous total mortality rate can also 

 be estimated from length composition data using 

 the expression derived in Appendix B of Beverton 

 and Holt (1956) based on the von Bertalanffy 

 (1938) growth parameters. This method is most 

 accurate when there is a rapid increase in length 

 with age over the important size range and a 

 minimum of overlap between the length distribu- 

 tions of adjacent age groups. Saila et al. (in press) 

 found that growth of male rock lobsters from the 

 Gisborne area during the first few years after re- 

 cruitment to the fishery (the important size range 

 in this study) is slow relative to earlier years as 

 described by an empirical growth curve. This 

 curve was considered to be a more realistic de- 

 scription of growth than the von Bertalanffy curve 

 at this stage of the species' life history. Moreover, 

 the great variability found in the individual 

 growth increments probably results in a high de- 

 gree of overlap between the lengths of adjacent age 

 classes. Thus, the conditions for the best use of the 

 Beverton and Holt method do not appear to be met, 

 and it was not applied to this species. 



Polymodal frequency distributions may also be 

 separated into their component groups using com- 

 puter techniques such as ENORMSEP (Extended 

 Normal Separator Program) ( Yong and Skillman 

 1975). An important advantage of this technique 

 over the method of Bhattacharya (1967) is that the 

 goodness of fit of the estimated component dis- 

 tributions to the original polymodal distribution 

 can be determined. However, the accuracy of both 

 techniques is reduced when the modes of the size- 

 frequency distribution are not well separated. The 

 size-frequency distributions analyzed in this 

 example do not exhibit any well-defined modes, so 

 ENORMSEP was not used as it is considered that 

 this technique would not improve the accuracy of 

 the estimates. 



The estimates of total mortality rate from the 



rate oftag returns using the method of Robson and 

 Chapman ( 1961) are much higher (even when ad- 

 justed for an 8- or 9-mo fishing season) than the 

 estimates derived from the size-frequency dis- 

 tributions. Morgan (1974a) suggested that the 

 western rock lobster, Pan!///r;/,s cygntis, caught by 

 pots have a higher probability of recapture by pots 

 than rock lobsters initially captured and released 

 by diving. He further suggests that rock lobsters 

 previously caught by pots, marked, and released, 

 have a greater probability of recapture by pots 

 than the probability of capture by pots of the total 

 population. All of the rock lobsters tagged and 

 released in this study were caught by pots, so they 

 may have been more vulnerable to capture by 

 commercial pot fishermen than the untagged in- 

 dividuals in the population. Moreover, due to the 

 nature of the fishing grounds, the tag and release 

 procedure was not conducted in a strictly random 

 manner, and the resulting distribution of tags in 

 the fishery may have led to a greater susceptibility 

 to capture for the tagged rock lobsters. 



The estimates of total mortality rate from the 

 tag returns also may be affected by changes in the 

 catchability of the rock lobsters. Laboratory ex- 

 periments indicate feeding activity is lowest for 

 individual J. edwardsii immediately prior to and 

 after molting (my unpubl. results). Morgan 

 (1974b) found a significant negative correlation 

 between premolt stage Dl animals and catchabil- 

 ity for the western rock lobster. Fluctuations in 

 the number of recaptures per unit relative effort 

 between August and April (Table 6) indicate that 

 catchability may vary considerably. Any decrease 

 in catchability during the molting period would 

 act to increase the estimate of the total mortality 

 rate. 



Other factors affecting estimates of the total 

 mortality rate (Type B errors, Ricker 1975) in- 

 clude: 1) tag loss, 2) extra mortality of tagged rock 

 lobsters, and 3) emigration of tagged rock lobsters 

 from the fishing area, with all three acting at a 

 steady, instantaneous rate throughout the exper- 

 iment. These factors all result in an overestimate 

 of the total mortality rate. 



Preliminary laboratory and field experiments 

 on the rates of initial tag loss and mortality due to 

 tagging indicate these effects are minimal (my 

 unpubl. data). However, rock lobsters are most 

 susceptible to tag loss and mortality due to the 

 presence of the tag while molting. Laboratory ex- 

 periments are being conducted to determine the 

 long-term rates oftag loss and mortality. 



478 



