822 



Fishery Bulletin 92(4), 1994 



Table 2 



Minimum detection limits (MDL), mean and ranges of estimated concentrations, and measurement error for 

 Nemadactylus macropterus, of the six elements that could be assayed reliably with a WD electron probe. The values 

 are based on a random subset of our data (n = 478 points), including numerous individuals and positions along the 

 scanned axis of points analyzed. Concentrations are given in ppm (by weight), except for Ca which is in percent of 

 the target mass. Note that values below the minimum detection limit are effectively zero. 'Minimum significant differ- 

 ence' is based on comparison of 'replicate' points in parallel life history scans (see text). CI = Confidence interval. 



Element 



MDL 



Minimum significant difference 



Mean (range) 

 concentrations 



Measurement error 

 (absolute, %) 



Mean (range) 



99% CI 



Ca 



Sr 



Na 



K 



S 



CI 



311 

 159 

 136 

 149 

 157 



38.8% (35.3-44.5) 



2240(1430-3860) 



3331 (2680-4240) 



729 (280-1630) 



421 (220-1220) 



255 (0-1230) 



±157(7%) 

 ±122(3.7%) 



± 72 ( 10%) 

 ± 76(18%) 

 ± 72(28%) 



210(12-964) 



160(10-450) 



77(0-320) 



80 (4-247) 



48(10-210) 



331 

 235 

 118 

 121 

 73 



10 



g> 



CD 



s 



-Q 10 



E 



Q. 



a. 



c 



o 



I 10 



c 



0) 



o 

 c 

 o 



o 



1 



Ca Sr S Hg Se Cd Mn Ni Br 

 Na K CI Ba Cu Fe Zn Pb 



Figure 3 



Mean (solid circle) and ranges (vertical line) of concen- 

 trations of the elements detected in the sagittae of 

 Nemadactylus macropterus by means of electron probe 

 (for Ca, Na, Sr, K, S, and CI) and proton probe (for 

 mercury (Hg), barium (Ba), selenium (Se), copper (Cu), 

 cadmium (Cd), iron (Fe), manganese (Mn), zinc (Zn), 

 nickel (Ni), lead (Pb), and bromine (Br)) microanaly- 

 ses. For proton probe methodology and results, see Sie 

 and Thresher (1992). Data are a compilation of >500 

 points across numerous individuals and positions along 

 the growth axis. The minimum detection limit (MDL) 

 for each element is indicated by the irregular horizon- 

 tal line and is based on the standard output of the re- 

 spective probe microanalyzers at our standard operat- 

 ing conditions (see Gunnetal., 1992; Sie and Thresher, 

 1992). The minimum concentration of CI is below the 

 detection limit of the electron-probe microanalyzer but 

 could not be determined with the more sensitive pro- 

 ton-probe microanalyzer because of our operating con- 

 ditions (see Sie and Thresher, 1992). 



centrations of 100-5000 ppm; and a variety of 'trace 

 elements' (e.g. iron, copper, and bromine) occur at 

 concentrations <10 ppm. Only the micro-constituents 

 and Ca can be measured accurately by WD-EPMA. 

 Absolute ranges of concentrations, measurement er- 

 ror (absolute and 95% confidence intervals), and 

 minimum detection limits (MDL's) for each of these 

 elements are given in Table 2. Measurement error is 

 inversely correlated with mean concentration, rang- 

 ing from 3.7% in Na to 28% in CI. Of the six ele- 

 ments measured, only CI occurred occasionally at less 

 than its respective MDL (157 ppm). Although the 

 microanalyzer reports values less than the MDL, 

 these values were considered noise and set equal to 

 zero for analyses of population structure. 



Life history scans for three fish chosen randomly 

 from the data set (Fig. 4) illustrate several points 

 typical of our data. First, all six elements vary onto- 

 genetically in concentration well in excess of the 

 uncertainty associated with measurement. Second, 

 concentrations at any given position correlate 

 strongly with those at neighboring points; for all six 

 elements, autocorrelations are highly significant at 

 scales <100 urn (e.g. Fig. 5), which suggests that in 

 N. macropterus this is the typical scale of ontoge- 

 netic variability in composition. Comparison of 'small 

 spot-closely spaced' analyses (6 urn beam diameter 

 spaced at 8 urn intervals) with 'large spot-widely 

 spaced' analyses ( 14 urn diameter at 16 urn spacing) 

 suggests that a sampling scale finer than our stan- 

 dard analysis reveals few, if any, major variations in 

 otolith composition that would not be detected at the 

 coarser sampling scale (Fig. 6). Third, absolute vari- 

 ability is highest for Sr, which can vary within speci- 

 mens over half an order of magnitude. However, rela- 

 tive variability is as high in S and CI; coefficients of 



