FISHERY BULLETIN: VOL. 82, NO. 2 



measured continuously with a lake level recorder 

 and calibrated every 2 wk with benchmark mea- 

 surements in Lake Nerka. To measure the effect of 

 light on the feeding success of char, we measured 

 incident solar radiation continuously with a 

 pyrheliometer. As a qualitative measure of light 

 intensity after sunset, we used a scale from to 4 

 where represented clear skies with relatively 

 high light intensity and 4 represented low over- 

 cast skies with relatively low light intensity. 



Data Analysis Procedures 



Functional Response of Char 



To determine the functional response of char 

 and the effects of other variables on the number of 

 consumed smolts, we grouped the data into 24-h 

 periods with 1200 h as the first hour of the day. The 

 number of smolts consumed per 24 h by an indi- 

 vidual char was calculated by the following 

 equation: 



N = S 



m 



where N -= number of smolts consumed/char per 



24 h 

 S == number of smolts observed in a char 



stomach 

 D - digestion time. 



Average digestion time (h) was determined from 

 data collected by Meacham and Clark (1979) and 

 calculated by the following curvilinear equation 

 (Fange and Grove 1979): 



Percent Smolt Mortality 



The average number of smolts consumed per 

 char, as described by the functional response, may 

 not represent the entire char population. Char 

 may migrate to the river to feed, then return to the 

 nearby lake environment for several days. Evi- 

 dence for this behavior stems from gill net catches 

 of char along the nearby lake shore and several 

 underwater observations of relatively few char in 

 the river during midday. Because the numerical 

 response of char is not known, 29c mortality curves 

 were developed from two hypothetical responses. 

 The first percent-mortality curve was based on the 

 assumption that the entire char population of 

 1,100 fish (Ruggerone 1981) fed each day (Type I 

 numerical response; Fig. 2A). The second curve 

 was based on the assumption that char immigrate 

 to the feeding area in response to smolt abun- 

 dance, thus a Type II numerical response was as- 

 sumed (Fig. 2A). 



Char Consumption of Smolts by Length 



A two-factor analysis of variance with replica- 

 tion (Zar 1974) was used to test for random con- 

 sumption of smolts by char. We divided the char 

 data into three time periods containing two levels 

 of char stomach fullness (full or less full) and three 

 sublevels of char length (295-445 mm; 446-470 

 mm; 471-502 mm). To concurrently compare the 

 length of smolts consumed by char with those 

 smolt lengths available in the migration, we cal- 

 culated the difference between average length of 

 smolts consumed and average length of smolts 

 available. This difference was utilized in each 

 level of analysis. 



InD = 4.892 - 0.143 (7 1. 



where D = digestion time 



T = temperature (°C). 



The functional response model based on multi- 

 ple regression analysis was developed with the 

 SPSS nonlinear program utilizing Marquardt's 

 method of least squares estimation (Marquardt 

 1963). Residual and partial residual analysis were 

 used to determine which independent variables 

 should be added and the shape of their partial 

 effect curve (Larsen and McCleary 1972; Draper 

 and Smith 1981). This method allows for analysis 

 of each new variable while including the effect of 

 previous variables. 



RESULTS 



Char Functional Response 



Nonlinear regression analysis indicated that 4 

 of the 12 variables tested affect the number of 

 smolts consumed/char per 24 h. The most impor- 

 tant of these variables was the number of smolts 

 migrating during the previous day's migration 

 (approximate partial F, P < 0.01). The next impor- 

 tant variable was the average weight of migrating 

 smolts (P < 0.01), then the number of smolts mi- 

 grating during the day of capture (P < 0.01), and 

 finally char length (P =£ 0.08). The amount of vari- 

 ability explained by all four variables was 59% 

 and the standard deviation was ±0.8 smolts 



404 



