"To Hit et al.: A method to improve size estimates of Theragra cha/cogramma and Pleurogrammus monopterygius 



503 



Results 



A relatively objective method to estimate the degree of 

 digestion of dominant structures of pollock and Atka 

 mackerel was derived by using defined criteria (Table 1) 

 and photo-reference material (Figs. 1 and 2). Condition- 

 specific digestion correction factors (and derived confi- 

 dence intervals) calculated for each structure augmented 

 our method of estimating size of prey from bones and 

 otoliths recovered in scats, as well as potentially from 

 bones and otoliths taken from stomach contents. 



Mean reduction (MR) in the size of pollock DENT 

 and QUAD in good condition and ANGU, HYPO, IN- 

 TE, OTO, and PHAR in fair condition were between 

 12.2-18.5%, and larger values were found for QUAD 

 (22.8%) and DENT (24.7%) in fair condition (Table 2). 

 Our overall 95% confidence intervals were generally 

 symmetrical and converted to a mean range of ±2.2% 

 (±0.5, SD) around MR values. Mean DCFs ranged be- 

 tween 1.14 and 1.33, and lower bounds of 95% CIs ex- 

 ceeded 1.11 in all instances, confirming that egested 

 structures of these condition categories were significant- 

 ly smaller than the size at which they were ingested 

 (Table 2). Partitioning errors showed that resampling of 

 egested structures was the major source of error (>73% 

 across structures) within the overall total. Our overall 

 95% CIs resulted in a maximum total error of ±1.7 cm 

 around an estimated mean of 40 cm for pollock. 



Mean reduction in the size of Atka mackerel struc- 

 tures varied more widely (3.3-26.3%), leading to DCFs 

 ranging between 1.03 and 1.36. QUAD in good condition 

 provided the smallest DCF, and DENT in fair condition 

 the largest. Overall, our 95% CIs converted to a mean 

 range of ±2.4% (±0.6) around MR values, and all lower 

 95% CI bounds exceeded 1.0 (Table 2). As seen when 

 errors were partitioned for pollock, errors owing to resa- 

 mpling of egested structures were the major source of 

 error (>83'7t across structures) within the overall total 

 for Atka mackerel. Our overall 95% CIs resulted in a 

 maximum total error of ±1.2 cm around an estimated 

 mean of 40 cm for Atka mackerel. 



With the exception of the two largest skeletal struc- 

 tures (DENT and QUAD, Table 2), some selected struc- 

 tures (INTE, HYPO, PHAR, ANGU, and OTO) occurred 

 in scats with no clear loss in size or loss of morphologi- 

 cal features related to digestion. For these five struc- 

 tures, we ascribed the condition category good and as- 

 signed a DCF of 1.0 (i.e., no correction for partial size 

 reduction due to digestion required). 



Of the 158 structures in our blind test, 141 (89.2%) 

 were assigned identical condition categories. Of the 

 remaining 17 structures, 11 (65%) were noted as be- 

 ing borderline between categories. Angulars (ANGU) 

 accounted for the majority (-60%) of all differences, 

 with all but one re-assigned in good condition as op- 

 posed to fair condition. On review, differences in as- 

 signing angulars were mainly the result of differences 

 in opinion on what constituted a well-defined and sharp 

 point (Fig. 1, Table 1). Clarification through the addi- 

 tional use of reference material (including both pristine 



structures and examples of each condition category) is 

 advised, particularly for angulars. Comparison of the 

 same 158 bones between two observers (D.T. and S.H.) 

 using the same structure reference collection resulted 

 in assigning more than 93% (147/158) of structures to 

 an identical category. 



The regression formula for estimating pollock mass 

 (M) from fork length (FL) estimates was best described 

 by using an exponential equation (M=0.0051 xFL 3n , 

 n = 981. r 2 = 0.987). 



Discussion 



The size of prey consumed by pinnipeds can usually be 

 reliably estimated from otoliths recovered in scats if 

 partial digestion is accounted for (Tollit et al., 1997). 

 However, otoliths from Steller sea lion scats are often 

 found in too few numbers, or are too digested or broken 

 to be useful (Sinclair and Zeppelin, 2002; Tollit et al., 

 2004, this issue). It was, therefore, necessary to use 

 alternative skeletal structures to estimate the size of 

 prey selected by Steller sea lions. Zeppelin et al. (2004, 

 this issue) documented good relationships (r 2 =0.78-0.99) 

 between the size of selected alternative structures and 

 fork length for pollock and Atka mackerel. However, 

 all skeletal structures are susceptible to digestion in 

 the stomach (our study, and Murie and Lavigne, 1986). 

 Thus, techniques are required to account for the degree 

 of digestion of alternative structures prior to estimating 

 prey size. 



Reductions in the size of otoliths during passage 

 through the digestive tract of pinnipeds have been 

 widely reported (e.g., da Silva and Neilson, 1985; Prime 

 and Hammond. 1987; Harvey, 1989; Tollit et al., 1997). 

 Similarly, we found significant reduction in the sizes of 

 all selected cranial structures from pollock and Atka 

 mackerel. Size reduction also showed great variability. 

 Relatively small structures were found with no obvious 

 loss in size due to digestion, but were also frequently 

 heavily eroded. 



The degree of digestion on different otoliths and 

 bones may be related to species, size of fish (Bowen, 

 2000). or even its shape, but seems to be random in any 

 one meal (Murie and Lavigne, 1986). Degree of diges- 

 tion likely depends on a range of factors such as meal 

 size, meal frequency, meal composition, and method of 

 consumption. In the face of these multiple factors we 

 feel our method for classifying the degree of digestion 

 into one of three condition categories is practical and 

 relatively objective. However, our technique does not 

 consider potential biases of enumeration associated 

 with smaller prey being more susceptible to complete 

 digestion than relatively larger prey, or of individual 

 fish being counted more than once if all multiple struc- 

 tures are used. Nevertheless, resolution to these biases 

 have been advocated (see Tollit et al., 1997; Laake et 

 al., 2002; Tollit et al., 2003; Tollit et al., 2004, this is- 

 sue). The category selections chosen with our criteria 

 showed good agreement among independent observers. 



