WATANABE and LO: PRODUCTION AND MORTALITY OF SAURY 



lute larval production in a limited area, which 

 can be covered with a high precision, is possible 

 and will be useful. 



The assumption that all the fish in a volume of 

 water strained by the net tow are captured in 

 the NIT period is not realistic for a variety of 

 reasons. The larval net used in this study has 

 bridles in front of the mouth, and to some extent 

 evasion can be expected due to turbulence in 

 front of the net. Visual evasion is possible even 

 in the NIT period because natural waters are not 

 totally dark due to moonlight and a variety of 

 bioluminescent organisms. 



Vertical distribution pattern of young sauries 

 has been investigated by towing nets on hori- 

 zontal strata at several depths (Okiyama 1965; 

 Parin 1967). However, the fraction of the young 

 saury population in the upper one meter avail- 

 able to our net is not known. 



The volume of water filtered by a tow may be 

 different among tows depending on wave condi- 

 tions, mesh clogging, etc. The standardization of 

 catch data by a volume of water filtered must 

 also be done to improve the accuracy of esti- 

 mates. In our study, we did not have a flow 

 meter with our net. Biases, resulted from these 

 problems, must be resolved before absolute lar- 

 val production can be obtained. 



Unaccounted variances due to the retention 

 correction factor and the duration computed 

 from gi'owth curve could lead to underestimation 

 of the variance of parameter estimates in the 

 mortality curve. On the other hand, the standard 

 error of both P,, and IMR in the current paper 

 may be higher than the estimates from a regres- 

 sion where each daily fish production estimate is 

 weighted by the inverse of its variance. The ef- 

 fect of the unaccounted variance caused by the 

 retention correction factor could be small be- 

 cause the retention rates were computed from a 

 large number of tows, reducing the standard 

 error of estimated retention to insignificant 

 levels. The variance of duration is unknown. 

 Theoretically it could be estimated from the 

 growth curve, but we did not compute it. A sim- 

 ulation study on evaluating the precision and 

 bias of mortality estimates for northern anchovy 

 (N.C.H.L., unpubl. data) indicated that the vari- 

 ation contributed from bias coiTection factors is 

 minimal. 



We employed an age-independent mortality 

 model in this paper. This does not necessarily 

 mean that the mortality of saury larvae is age- 

 independent. Because of biased catch data due to 

 mesh extrusion in the smallest two size classes. 



we could not use these two values. This might 

 have had some effect on choosing a mortality 

 model. For further studies we have devised a 

 cylinder-conical larval net for saury that is con- 

 structed with one type of mesh (0.53 mm). The 

 body depth at the pectoral fins in early post- 

 yolk-sac larvae is 0.6-0.7 mm, so the new net is 

 expected to retain larvae of the smallest size 

 classes. Thus, the net will enable us to get an 

 unbiased sample of early larvae, and the use of a 

 data set of all size classes including 7.5 and 12.5 

 mm classes might cause a shift of the mortality 

 model from age-independent to age-dependent 

 type. 



Because mortality is a necessary piece of infor- 

 mation for computing mean age for each size 

 group, we used the ages corresponding to the 

 midpoints of the size classes for the mortality 

 computation instead of mean ages. It is accept- 

 able to use the midpoint to convert size to age 

 when the mortality information is unavailable, 

 when mortality is low, or when the size interval 

 is small. Bias resulted from the use of the mid- 

 point was large for the first size class: 2.4 days 

 with mortality correction assuming the IMR = 

 0.08 and 3.2 days using the midpoint 7.7 mm. 

 However, the first two size classes were not 

 included in the mortality computation, so bias 

 caused by using the midpoints was minimal. 



The high production values in large juvenile 

 classes (Table 2, Figs. 4, 5) may be due to an 

 underestimation of the durations of these size 

 classes. Year-to-year variations of growth may 

 be large enough to cause considerable differ- 

 ences in durations. The standing stock of larvae 

 in a size class is influenced by the duration of 

 growth through that size class. Because an ac- 

 curate mortality estimate depends on accurate 

 grovrth rate estimates, calculation of the growth 

 rate every year, or at least every time the fish 

 size composition changes, is necessary to obtain 

 accurate mortality rates for individual years. 



As shown for the northern anchovy, direct 

 estimates of spawning biomass from an ichthyo- 

 plankton survey can be obtained (Lasker 1985). 

 This method can be applied to other fishes that 

 produce pelagic eggs. However, the saury pro- 

 duces adhesive eggs and quantitative sampling 

 of them is difficult. Larval census is so far the 

 best index for spawning biomass of the saury. 

 Even if it were possible to obtain quantitative 

 samples of eggs, uncertainties in embryonic mor- 

 tality rates may rule out use of egg production. 

 The saury has a long incubation time of about 2 

 weeks under 13°C (Yusa 1960), and slight differ- 



611 



