Dorn: Environmental covanates of Merlucaus productus growth 



271 



fishery would decrease, and this would tend to decrease 

 the length-at-age in the catch in the following year from 

 what it would have been otherwise. From 1978 to 1982, 

 the fraction of the Pacific whiting catch taken north 

 of Cape Falcon (near the mouth of the Columbia River 

 at lat. 46°45'N) increased from 10% to 80%, and was 

 60% in 1988 (Fig. 6). The lack of fit of the growth- 

 increment regression model with respect to year in 

 Figure 5 may have been a result of this northward shift 

 in fishing mortality. 



However, it is difficult to predict the long-term ef- 

 fects of a shift in the geographic pattern of exploita- 

 tion on length-at-age, because little is known about the 

 extent of mixing from one year to the next of fish 

 migrating from different regions. Without mixing be- 

 tween regions, an increase in fishing in the northern 

 part of the range would reduce the abundance of larger 

 individuals of an age-group, reducing the overall pop- 

 ulation length-at-age, while length-at-age of the south- 

 ern fish would be unaffected. A more likely hypothesis 

 is that some inter-regional mixing occurs from year to 

 year. In this case, the length-at-age in all regions would 

 decrease, though the magnitude of the decrease should 

 be greatest in the north where the higher fishing oc- 

 curred. Ultimately, this would tend to reduce latitudinal 

 variation in length-at-age. 



To support this hypothesis, there is some 

 evidence of a change in the degree of latitu- 

 dinal segregation by size of Pacific whiting. 

 Figure 7 shows the coefficients for a region- 

 year interaction for a length ANOVA with 

 main effects being age, year, region, sex, and 

 season. The absence of interaction between 

 year and region would be identified by parallel 

 year effect lines for each region, and would in- 

 dicate that size-specific migratory pattern has 

 remained constant. From 1978 to 1985, the 

 region-year interaction does not appear promi- 

 nent. After 1985, however, the lengths-at-age 

 in the three regions become much closer 

 together; in particular, the lengths of the fish 

 in the VNC region, instead of being 1-2 cm 

 larger than the fish in the other regions, are 

 the same size or smaller. 



Recently, Smith et al. (1990) examined the 

 length-at-age data from the fishery for Pacific 

 whiting in the Canadian waters over the same 

 years examined by this paper. They used 

 a generalized form of the von Bertalanffy 

 growth model that makes length-at-age a func- 

 tion of length-at-age in the previous year, plus 

 environmental covariates modeled in different 

 ways according to a hypothesized mechanism 

 by which the environmental covariate affects 

 growth or apparent growth. Significant covari- 

 ates in their model were population biomass, a suite 

 of oceanographic variables measuring the strength of 

 southward advection of water from the Alaskan Sub- 

 arctic Gyre, and several variables that model size- 

 selective mortality. Since fish younger than age 5 are 

 not common in the Canadian samples, their analysis 

 could not examine the sources of growth variability of 

 the younger fish. Consequently, the analysis presented 

 here on the environmental covariates of Pacific whiting 

 growth in U.S. waters is a necessary complement to 

 the paper by Smith et al. (1990). For example, an in- 

 verse relationship between temperature and growth, 

 which is most pronounced for the younger fish, was not 

 detected by Smith et al. (1990), and can partly account 

 for the fact that the fish currently recruiting to Cana- 

 dian waters at age 5 are much smaller than those 

 recruiting in the late 1970's. 



A major contention of Smith et al. (1990) is that ex- 

 pansion of the Canadian fishery is largely responsible 

 for the decline in length-at-age observed since 1976. 

 They used the ratio of the Canadian catch (in biomass) 

 to the total population biomass during the current year 

 as a covariate in their nonlinear regression model, a 

 phenomenological approach that sidesteps the need to 

 model the population dynamics. Although the mono- 

 tonically increasing Canadian catch of Pacific whiting 



