Baba et al.: An environmentally based growth model for |uvenile Corbicula japonica 



21 



Table 2 



95% confidence limits of location and scale parameters at the first sampling and coefficients of the best model constructed based 

 on the largest extreme value distribution (models 4.1 in Table 1) estimated by profile likelihood method. dRIRL = daily relative 

 increase rate of location parameter. dRIRS = daily relative increase rate of scale parameter. Temp. = water temperature, WF = 

 water fluorescence, Sal. = salinity, Turb. = turbidity. 



Parameters 



at 1st 



sampling 



Max. 

 dRIRL 



Age categorization 



Environmental factors 



Expressing 



relationship 



between dRIRS 



and dRIRL 



A, 

 a, 



a. 



Temp. 

 ft 



WF 

 ft 



Sal. 



ft 



Turb. 



ft 



Lower 95 % 

 Upper 95 % 



0.294 

 0.304 



0.039 

 0.045 



0.010 -26.6' 



0.013 -11.5' 



-14.6 

 -6.4 



0.41 

 1.00 



0.27 

 0.64 



0.0027 

 0.0039 



0.734 

 0.793 



1 One common coefficient for the two categorical variables. 



13 August. Then it sharply decreased to 3 ind/m 3 on 19 

 August (Baba et al., 1999). Such a pattern of larval-density 

 fluctuation might have caused the asymmetric distribution 

 of shell lengths of the settled juveniles. Another possible 

 factor that influenced the shapes of the shell length distri- 

 butions and the relationship between dRIRL and dRIRS 

 is size-dependent mortality, e.g. predations and fisheries. 

 Size-dependent mortality has been reported in several 

 marine bivalves (e.g. Nakaoka, 1996). Potential predators 

 of C.japonica are fishes, such as Japanese dace (Tribolo- 

 don hakonensis) (also known as big-scaled Pacific redfin, 

 FAO), Pacific redfin (Tribolodon brandtii), common carp 

 (Cyprinus carpio), and the So-iny mullet (Liza haemato- 

 cheila ) (Kawasaki 4 ). In our study, the size-dependent mor- 

 tality was negligible because the range of the shell lengths 

 observed in this study was very narrow. 



The shape of the distribution to describe a single cohort 

 should be determined from the data. In contrast, single 

 cohorts are usually separated from multicohort data by as- 

 suming a normal distribution of lengths in a single cohort 

 (e.g. Fournier and Sibert, 1990). Therefore, it is possible 

 that multicohort analysis done without selection of an 

 adequate distribution to describe a single cohort causes 

 substantial bias in estimations of various stock features 

 of animal populations, such as age composition, growth, 

 mortality, and recruitment. In our preliminary analyses, 

 we also tested smallest extreme value distribution, inverse 

 Gaussian distribution, and lognormal distribution. The in- 

 verse Gaussian distribution was the best for two samples; 

 the lognormal distribution, was the best for two samples; 

 the largest extreme value distribution was the best for ten 

 samples. Therefore, it is reasonable to select the largest ex- 

 treme value distribution. We selected a single distribution 



for our analyses, otherwise a discontinuous point would 

 have appeared in the growth curve. 



Relatively large confidence intervals were obtained in 

 the coefficients of the linear component of Equation 6, i.e. 

 a , and /3 ; , (Table 2). The relatively large confidence inter- 

 vals may indicate that the number of estimated coefficients 

 is somewhat larger than the number of samplings. There- 

 fore, to estimate these coefficients more precisely, we may 

 need to investigate more cohorts spawned in other years 

 in future investigations. 



Growth of C. japonica 



We identified extremely slow growth in C. japonica juve- 

 niles, which grew to a modal shell length of 0.7 mm during 

 the first year in Lake Abashiri, which lies at 43.7°N. Spats 

 of C. japonica collected from 1992 to 1997 in Lake Shinji, 

 which lies at 35.5°N, grew to a mean shell length of 6.7 

 mm in natural conditions by the first winter (Yamane et 

 al. 2 ). Using environmental factors measured in Lake Shinji 

 from 1990 to 1998 at monthly intervals (Seike 5 ), we simu- 

 lated the growth of C. japonica with model 4.1. Corbicula 

 japonica grew to a mean shell length of 1.4 mm (standard 

 error, 0.37 ) by the first winter in the simulations. Therefore, 

 the large difference in juvenile growth between the two 

 habitats cannot be explained by environmental differences 

 because the results of the simulation were apparently an 

 underestimate. We think that the extremely slow growth 

 of the juveniles (prolonged phase of meiobenthic develop- 

 ment ) in Lake Abashiri is probably a geographical varia- 

 tion, which is genetically determined, within C. japonica. 

 However, there remains a possibility that the juvenile 

 growth differences depend on other environmental factors 

 not measured in this study. Therefore, the geographical 



4 Kawasaki, K. 1997. Lagoon structure and fish produc- 

 tion in Ogawara-ko Lagoon. /;; Final reports on fisheries in 

 Ogawara-ko Lagoon (Tohoku Construction Corporation ed.), 

 p. 4-33. Unpubl. rep. Construction Office for Takasegawa 

 General Development of Tohoku Regional Construction Bureau, 

 3 Ishido, Hachinohe, Aomori 039-1165, Japan. 



6 Seike, Y. 1990-98. Gobiusu: monthly report of water quality 

 in Lake Shinji and Lake Nakaumi. Unpubl. rep. Faculty of 

 Science and Engineering. Shimane University, 1060 Nishi- 

 kawatsu, Matsue, Shimane 690-0S23, Japan. 



