Thresher et al.: Otolith analysis of Nemadactylus macropterus 



827 



J[ 



h 



Na 

 x = 3300 

 P N >0.5 



20 



Sr 

 x = 2230 



P N > 01 



15 



30 



2700 



4200 1600 



3100 55 



JO 



A 



s 



x = 450 

 p <0.01 



25 





_ Cl 

 x = 360 

 p < 0.001 



c 

 3 

 cr 



? 

 1900 O 



3 

 Q. 



20 | 



Cffl^ 



c 

 BI 



38% 



46% 300 1050 



Concentration 



1500 



Figure 9 



Frequency distributions of concentrations of the six detected elements (Ca, Na, 

 Sr, K, S, and Cl) at point 2 for 64 adult Nemadactylus macropterus pooled across 

 sites in SE Australia, /^probability the distribution is normal. Concentrations 

 are reported in units of ppm by weight for all elements other than Ca, which is 

 reported as percent by weight. 



for individual fish is greater for the single-point data 

 than for the filtered data, but the patterns of regional 

 groupings and the relationships between the concen- 

 trations of elements are much the same for the two 

 data sets. Victorian and NSW samples appear to 

 group based on both Na/Cl and Sr/Ca ratios. The two 

 Tasmanian samples group with specimens from the 

 Bight on the basis of Na/Cl ratios but appear to dif- 

 fer based on mean Sr/Ca ratios. 



Linear discriminant function analysis (LDFA), for 

 a 3-group discrimination, produces similar results 

 for both single-point and filtered data. For both data 

 sets, the three groups are statistically separable but 

 at a relatively low rate of successful classifications. 

 For the single-point data, 66% of individuals were 

 accurately classified into their three respective 

 'source populations'; for the filtered data, the suc- 

 cess rate increases to 78%. The relatively poor sepa- 

 ration is due, in part, to an overlap of the three groups 

 in discriminant space and, in part, to a few individu- 

 als located in discriminant space well outside the 

 areas defined by most individuals collected at the 

 same place and time. 



For the filtered data, discriminant analysis devel- 

 oped five discriminant functions to classify the six 



sites. However, only the first three are significant 

 (P<0.05), and of these there is a large difference be- 

 tween the first two functions (both at P<0.01) and 

 the third (P=0.02). Examination of the canonical load- 

 ings indicates that discriminant function 1 is corre- 

 lated with Na, K, and Cl concentrations, and hence 

 represents mainly the initial separation of sites along 

 the Na/Cl axis indicated in Figure 11. The second 

 discriminant function loads heavily only on Sr, 

 whereas the third is mainly a K residual from the 

 first discriminant function. Step-down procedures, 

 in which sites are sequentially pooled, raises the con- 

 tribution of Ca to the second discriminant function, 

 identifying it with the Sr/Ca axis in Figure 11. The 

 nature of the site separations is indicated in Figure 

 13. Function 1 separates the two Tasmanian and the 

 Bight ( GAB ) samples from the two Victorian and the 

 NSW samples; function 2 distinguishes weakly be- 

 tween the GAB sample and the remainder; and func- 

 tion 3 separates the east and west coast Tasmanian 

 samples. The remaining two functions do not clearly 

 distinguish among any sites. The primacy of the first 

 three functions remains in a step-down procedure, 

 as the sites are sequentially pooled based on their 

 degree of overlap. The final step, at which all func- 



