204 



Fishery Bulletin 91 12). 1993 



Table 6 



Standardized canonical coefficients from the Canonical Dis- 

 criminant Analysis of density of fishes over each method by 

 night combination, from samples taken off Lizard Island. 

 Great Barrier Reef on 3, 5, and 6 December 1986. Data were 

 lni x+ 1 1 transformed. 



* Consistently high values in total, between and within 

 canonical structure. These variables contribute significantly 

 to the discriminatory power of the canonical variate. 



Data from all five nights provided more information 

 on patterns of temporal change for some taxa (Fig. 7). 

 We focused on the comparative ability of the different 

 methods to detect changes over time in numbers of the 

 larger (>6mm) individuals of some families because 

 we wished to know the best methods for identifying 

 temporal pulses of large larvae and pelagic juveniles 

 of reef fishes. Large pomacentrids and mullids serve 

 as appropriate examples. Although absolute numbers 

 of fishes taken by nets and aggregation devices could 

 not be directly compared, temporal changes in pat- 

 terns of density could be evaluated among these meth- 

 ods. Comparisons were made using all methods, al- 

 though bongo net data were available for the nights of 

 3, 5, and 6 December only. 



Data from the two aggregation devices indicated that 

 large pomacentrids increased in density from the 2nd 

 to a peak on the 5th, and decreased over the 6th and 

 7th (Fig. 7). This pattern was not present in the data 

 from nets, each of which provided a different temporal 

 pattern of density. 



\J*y purse seines 



\j|jl}/ bongo nets 



neuston net3 

 •V Tucker trowls 



+ ATHERINIDAE 

 6 



+ APOGONIDAE 



10 



Figure 6 



Results of Canonical Discriminant Analysis of density data tnumbers/lOOOm') for apogonids, atherinids, gobiids, lethrinids. mullids, and 

 pomacentrids taken by four net types on the nights of 3, 5, and 6 December 1986 off Lizard Island. Great Barrier Reef. Factors analyzed 

 were net type and night of sampling. Canonical variates 1 and 2 are displayed. Numbers superimposed on circles refer to the day of 

 sample. 



