Effects of microprobe precision 

 on hypotheses related to 

 otolith Sr:Ca ratios 



Christopher L. Toole 



Department of Fisheries and Wildlife. Oregon State University 



104 Nash Hall. Corvallis, Oregon 97331-3803 



Present address: Environmental and Technical Services Division 

 National Marine Fisheries Service. NOAA 

 911 NE 1 1th Street. Suite 620. Portland. Oregon 97232 



Roger L. Nielsen 



College of Oceanography. Oregon State University 

 Ocean Administration 104. Corvallis. Oregon 97331-5503 



Several recent studies have used 

 the electron microprobe to infer 

 environmental temperature at the 

 time of otolith formation from the 

 concentration ratio of strontium 

 and calcium. Sr/Ca ratios of otoliths 

 from fish held at constant tempera- 

 ture or collected at known temper- 

 ature were examined using atomic 

 absorption spectrophotometry 

 (Radtke 1984, 1989) or wavelength 

 dispersive electron microprobe an- 

 alysis (Townsend et al. 1989, Kalish 

 1989, Radtke et al. 1990). These 

 studies, with the exception of Kalish 

 (1989), concluded that there is a 

 negative linear relation between en- 

 vironmental temperature and oto- 

 lith Sr/Ca ratio. This relationship, 

 coupled with assignment of age to 

 each microprobe sample site, has 

 been used to infer the relative tem- 

 perature histories of wild-caught 

 fish (Radtke 1984,1987,1989; 

 Radtke and Targett 1984; Radtke 

 and Morales-Nin 1989; Townsend et 

 al. 1989; Radtke et al. 1990). The 

 most ambitious application of the 

 method used otolith Sr/Ca ratios to 

 contrast the calculated temperature 

 histories of different subpopula- 

 tions of larval herring in the Gulf of 

 Maine (Townsend et al. 1989). 



Using the electron microprobe to 

 calculate individual fish tempera- 

 ture histories from otolith Sr/Ca 



ratios is potentially a useful tech- 

 nique for fisheries biologists. How- 

 ever, precision of back-calcuJated 

 temperature estimates should be 

 examined in greater detail. Previ- 

 ous studies do not explicitly state 

 confidence limits for mean re- 

 sponses or prediction limits for new 

 observations. The scatter of points 

 in Radtke (1989), Townsend et al. 

 (1989), and Radtke et al. (1990) 

 suggest that widths of 95% pre- 

 diction limits may be on the order 

 of one to several °C for most levels 

 of Sr/Ca examined. While this 

 might be acceptable for studies of 

 fish which are exposed to wide 

 variations in environmental tem- 

 perature, it is of less use for species 

 which experience more subtle tem- 

 perature changes. 



Future validation experiments 

 may improve the predictive capabil- 

 ities of the Sr/Ca vs. temperature 

 relationship by examining effects 

 of other variables. For instance, 

 the regression model might be ex- 

 panded to include growth rate (Ka- 

 lish 1989) and some measure of 

 physiological stress (Townsend et 

 al. 1989), since these also appear to 

 influence the Sr/Ca ratio. 



However, one component of the 

 variation not likely to change in fu- 

 ture experiments employing the 

 electron microprobe is the model er- 



ror term associated with measure- 

 ment. Usually measurement error 

 is considered insignificant in rela- 

 tion to other sources of variation 

 and is incorporated into the total 

 error term: 



Y = a + b*X + exotai 



where Gxotal = ^Measurement + ^Other ■ 



Measurement error can be thought 

 of as a lower bound to the varia- 

 tion associated with the regression 

 model when other sources of error 

 are minimized. 



We suspect that measurement 

 error may be nontrivial when de- 

 riving Sr/Ca vs. temperature rela- 

 tionships. Sr/Ca ratios associated 

 with a 1°C change in environmen- 

 tal temperature were approximate- 

 ly 0.00013-0.00036 in previous 

 studies (Table 1). It is difficult to 

 evaluate the significance of these 

 small values without more informa- 

 tion on the analytical precision of Ca 

 and Sr detection in fish otoliths 

 using the electron microprobe. Of 

 the studies cited above, only Kalish 

 (1989) reported analytical precision 

 for representative values of Sr and 

 Ca. In that study, measurement er- 

 ror associated with Sr was 3.5% and 

 that associated with Ca was 0.5% 

 for an Sr/Ca ratio of 0.002. 



One purpose of the present study 

 was to examine the precision asso- 

 ciated with measuring Sr/Ca ratios 

 in fish otoliths, and to demonstrate 

 how this error affects temperature 

 estimates derived from published 

 regressions. Our approach was to 

 intensively sample one otolith from 

 one fish at three beam-power den- 

 sities and four counting times. By 

 using one otolith, between-fish ef- 

 fects could be ignored. Within-fish 

 Sr/Ca effects were minimized by 

 referencing samples to the same 

 growth zones, leaving the different 

 analytical techniques as the primary 

 source of variation. 



Manuscript accepted 15 January 1992. 

 Fishery Bulletin, U.S. 90:421-427 (1992). 



421 



