FISHERY BULLETIN: VOL. 81, NO. 2 



from each other in mean pod size estimated (1.842, 

 1.829, t= 1.1442 < t.0.05) and, consequently, the dif- 

 ference of this year's data from others cannot be 

 credited to one aberrant observer. 



There was a significant increase in mean group size 

 as a function of distance from shore (Fig. 4) (F = 

 97.28 > F 5f 23i). A significant linear increase in the 

 pooled data (Fig. 4) was also noted in 10 of the 13 

 individual years. In the remaining 3-yr data ( 1968-69, 

 1972-73, 1978-79), the average pod size peaked at 

 about 0.6-0.9 km (1-1.5 mi) from shore, and de- 

 creased thereafter. This may be a real between-year 

 difference in whale behavior, but is more likely a 

 function of the varying abilities of the observers 

 themselves. 



There are highly significant differences between 

 years in the frequency of observations recorded 

 within offshore distance intervals (x 2 = 2,340, df = 

 24). For this analysis, a pooled three-interval dis- 

 tribution was used in light of the observer bias tests 

 discussed above. Within both the Yankee Point loca- 

 tion subset of years and the Granite Canyon subset 

 there also exists significant heterogeneity in the 

 offshore distribution (x 2 = 1,077, df= 14;x 2 = 1,025, 

 df= 8, respectively). Given this, a difference between 

 locations pooled overyears (x 2 = 239, df = 2) is not 

 surprising and also not particularly meaningful.Con- 

 sequently, given the range of interyear variation, we 

 cannot adequately test for interlocation differences 

 in the migratory corridor and therefore have applied 

 distance estimation corrections equally to data from 

 both locations. 



Within each year, the distribution of distance es- 

 timates was tested for a within-season change, since 

 our verification experiments were conducted during 

 roughly the middle third of the migration. For this, 

 the data were divided into early (10-29 December), 

 mid (30 December-18 January), and late (19 Janu- 

 ary-6 February) time periods. As with the first 1 1-yr 

 data (Reilly et al. 1980), the 1978-79 and 1979-80 

 distributions have no seasonal differences indicated 

 by contingency table analysis (x 2 = 8.54, 7.13, < 

 X% 0.05)1 but do have significantly different mean dis- 

 tance observations (F = 16.34,26.91 > F 2 xooh ). Con- 

 sequently, as with the first 1 1 yr, only data from the 

 middle third of the migration were used for com- 

 parison with aerial results in Equation (3). 



No significant period differences were indicated for 

 the 1978-79 and 1979-80 censuses, in the ANOVA 

 testing for effects on numbers of whales recorded per 

 5-h shift, from variation between observers and from 

 period (morning or afternoon). Similar results were 

 obtained in the comparison of observers within each 

 year (F = 1.242, 2.003, F 1I18 ). The data were 



0.5 



_l_ 



_J_ 



_L 



JL 



0.25 0.50 0.75 1.00 1.25 1.50 1.75 



DISTANCE FROM SHORE (nautical miles) 



2.00 



FIGURE 4. — Mean pod size estimates by distance from shore, with 

 95% confidence limits, from 13 annual gray whale census, 1967-68 

 through 1979-80, n = 23,678 observations. 



therefore considered homogeneous for pooling over 

 these factors. 



The results from 1978-79 are somewhat different 

 than the results from the first 11 yr, in the rate of 

 whales recorded per hour of day. The mean counts 

 show significant differences in an ANOVA (F= 3.717 



>F 



9, », 0.05 



) which are due to the depressed value for 



0700-0800 h (Fig. 5). Multiple range tests (Duncan's, 

 Student-Newman-Kuel's) indicate that the hourly 

 means (other than that for 0700-0800 h) are homo- 

 geneous. 



10 



i 6 



O 

 U 



z 



< 4 

 LU 



2 - 



A t 



,, f 4 

 -J A A ® J, 



10 



12 



14 



16 



18 



HOUR OF DAY 



Figure 5.— Mean count of gray whales passing Granite Canyon Sta- 

 tion by hour of day, with 95% confidence limits. 1978-79 and 1979- 

 80 pooled, n = 120 d. 



Modeling of Migratory Timing 



Table 3 lists the parameters of the gamma distribu- 

 tion as calculated for all 13 yr, along with the mean 

 days (— a p) and standard deviations ( Vcr /?) of the 

 annual migrations. As previously discussed, the 



274 



