Maxwell et al.: Fishery dynamics of Loligo opalescens 



665 



One measurement was made at each station 

 at sea surface and at 75 meters depth during 

 April (n = 8 for both depths). 



Results 



Ground-truthing: aerial observations 

 of boat activity 



Nonsquid vessels used weak lights (i.e., much 

 less than 30,000 watts), which did not show 

 in the satellite images. On average, 23 squid 

 fishing vessels were observed each night by 

 the aerial surveys (range = 0-64 vessels, n=26 

 nights). The 20:00-midnight observation 

 period was the peak time for attraction of 

 squid by the light boats. Although the squid 

 vessels did change location during this time, 

 they typically left their lights running to 

 continue searching for squid. The number of 

 squid vessels explained much of the varia- 

 tion in detected light pixels; proportion lunar 

 phase failed to enter the analysis as a signifi- 

 cant variable (Table 1). Detected light pixels increased 

 with the number of squid vessels (Fig. 2). 



The regression analysis yields the following simpli- 

 fied equation: 



log 10 (p, +l) = 1.25xlog 10 (x, +1), (1) 



where x t = observed number of squid vessels; and 

 p t = detected light pixels for night t. 



We used inverse prediction to estimate the number 

 of squid vessels for each satellite night (E t \ in the 

 1992-2000 period (Zar, 1984). The estimated number 

 of squid vessels was found by the equation 



Table 1 



Multiple stepwise (forward selection) regression of de- 

 tected light pixels on squid fishing vessels (transformed: 

 x'=log 10 (x+l) and proportion lunar phase (transformed: 

 .r'=arcsin( Vr). r 2 =0.64; ANOVA: F l 24 = 42.66, P<0.0001. 



Variable 



Coefficient ±SE 



Squid fishing vessels 

 INTERCEPT 

 Proportion lunar phase 



1.25 ±0.19 

 0.07 ±0.24 

 not entered 



<0.0001 

 >0.75 

 not entered 



4=io"«- ( ft +imB -i= 1J »/ft+i-i- 



(2) 



The ground sample distance of the satellite data is 

 2.7 km, which means that multiple squid vessels may 

 potentially fit into one pixel of detected light. This could 

 result in an underestimation of effort. The severity of 

 this problem can be assessed by examining the coeffi- 

 cient of the simple linear regression of log-transformed 

 variables represented by Equation 1. One of four sce- 

 narios is possible: 1) boats are not aggregated (coef- 

 ficient^), 2) boats are aggregated regardless of the 

 number of boats on the water (coefficients), 3) boats 

 are aggregated only when many boats are on the water 

 (coefficient<l), or 4) boats are aggregated only when few 

 boats are on the water (coefficient>l). The coefficient 

 in Equation 1 is 1.25, which fails to significantly differ 

 from 1.00 (f-test for regression coefficient: t = 1.305, 

 j3 = 1, df = 24, P > 0.2, two-tailed; power < 0.5, retrospec- 

 tively calculated; Zar, 1984). This result suggests that 

 very little clumping of the boats occurred (scenario 1), 



or that the degree of clumping was independent of the 

 number of boats on the water (scenario 2). Although the 

 statistical power of this ^-test is not high (power<0.5), 

 we conclude that the data provide more support for sce- 

 narios 1 and 2 over scenarios 3 and 4. Either scenario, 

 1 or 2, allows for a comparison of the relative values of 

 estimated effort and LPUE within a time series. 



Fishery characteristics, 1992-2000 



A composite satellite image of all squid fishing activity 

 in the Southern California Bight during the 1992-2000 

 study period revealed major concentrations of effort off 

 the Channel Islands, especially Santa Rosa, Santa Cruz, 

 Anacapa, and Santa Catalina (Fig. 1A). Squid fishing 

 occurs close to the island shores and is bounded by the 

 100-m contour. During the study period, 379.2 billion 

 kg of squid were landed in the bight: 341.2 billion from 

 blocks 651-896 (Fig. IB), and the remainder from the 

 large blocks 1032-1035. The main areas of fishing activ- 

 ity, as indicated by satellite, are consistent with the 

 blocks of high catch (Fig. IB). We note that blocks 682 



