McKinnon: Feeding habits of Lagenorhynchus obscurus 



571 



Historia Natural de la Universidad Nacional Mayor 

 de San Marcos, Lima), and with the assistance of S. 

 Candela (University of Miami, Florida). 



Counts and measurements of undigested hard 

 parts of prey allowed calculation of several measures 

 of the relative importance of each prey species. The 

 simplest measure, "percent frequency of occurrence", 

 was defined as 100 multiplied by the number of stom- 

 achs in which a prey species was present/the total 

 number of stomachs in the sample, excluding empty 

 stomachs. "Percent total numbers," was defined as 

 100 multiplied by the number of individuals of a spe- 

 cies of prey/the sum of individuals for all prey spe- 

 cies (Frost and Lowry, 1980). The number of indi- 

 viduals of each prey species in a sample was esti- 

 mated by dividing the count of its otoliths (for fish) 

 or squid beaks (for squid) by two (Frost and Lowry, 

 1980). 



Lengths and weights of consumed fish were esti- 

 mated by using regressions involving fish length and 

 otolith length, or fish weight and either fish length 

 or otolith length. Calculations for anchoveta, 

 Engraulis ringens. Pacific Sardine, Sardinops sagax, 

 hake, Merluccius gayi, and horse mackerel, Track- 

 urus symmetricus, followed Chirinos and Chuman 

 (1968), Samame (1977), McKinnon (1988), and 

 Hawes (1983), respectively, except that length was 

 estimated for T. symmetricus by using L=4.37xW 1/3 , 

 and weights of E. ringens and S. sagax were esti- 

 mated by using W=0.007xL 3 and W=0.015xL 3 , where 

 W=weight (g) and L=total length (cm). 1 



For each fish species, ten randomly selected 

 otoliths from each stomach were measured (Murie, 

 1984). If fewer than ten suitable otoliths were 

 present, all those available were utilized. Only 

 otoliths with minimal degradation were measured. 

 Degradation was apparent from a loss of detail, par- 

 ticularly the loss or reduction of spines and lobulations 

 along the edges of the otoliths (Frost and Lowry, 1986). 



Squid mantle lengths were estimated for each 

 squid species by using linear regressions of mantle 

 length on rostrum length and squid weights from 

 regressions of log e weight on log rostrum length 

 (Wolff, 1984). Regressions were not available for 

 patagonian squid, Loligo gahi, so regression equa- 

 tions for Loligo opalescens, a closely related species, 

 were used (Wolff, 1984). For each squid species, ten 

 randomly selected beaks were measured from each 

 stomach, unless fewer than ten were present, in 

 which case all were utilized. 



A mean individual weight (MIW) was calculated 

 for each prey species in each stomach and then used 



1 Pauly, D. International Center for Living Aquatic Resources 

 Management, Manila. Personal commun., 1985. 



in estimating the percent weight contribution of each 

 prey species to the dusky dolphin's diet. The MIW 

 was usually the mean of the regression-estimated 

 weights of individuals of a given prey species in a 

 particular stomach, unless all hard parts were too 

 degraded to permit reliable measurement, in which 

 case an overall MIW, the mean of all regression-esti- 

 mated lengths of that species in all stomachs with 

 measurable hard parts, was employed. For ancho- 

 veta, however, enough measurable otoliths were 

 available from stomach samples to permit statisti- 

 cal analyses by year and season of capture. Overall 

 MIW's for anchoveta were therefore calculated for 

 each group of stomachs within which analyses re- 

 vealed no significant differences (for example, the 

 summer of 1985; see Results). 



The total weight of each species of squid or fish 

 present in each dolphin stomach was estimated by 

 multiplying the number of individuals present by the 

 appropriate MIW value. The percent weight of each 

 prey species in the dusky dolphin's diet was calcu- 

 lated by using weights summed over all stomachs, 

 as 100 multiplied by the total weight of each prey 

 species/the total weight of all prey present. Species for 

 which regression-estimates of length and weight were 

 not available were excluded from these calculations. 



The percent gross energy contribution of each spe- 

 cies was defined as 100 multiplied by gross energy 

 of the prey species/summed gross energy of all prey 

 consumed, for all stomach content samples. The gross 

 energy available from a prey species is the caloric 

 density (kcal-g -1 )xweight (g) consumed. Caloric den- 

 sity (CD) values were obtained from the literature 

 for each prey species, either directly from bomb-calo- 

 rimetric analyses or indirectly from data on proxi- 

 mate composition, by using CD's for fat, carbohy- 

 drate, and protein of 9.4, 4.15, and 5.65 kcalg -1 , re- 

 spectively (Pike and Brown, 1984). 



By using published data from non-El Nino years 

 only, CD values for anchoveta were calculated as 

 1.589 kcalg -1 for the summer and 1.548 kcalg -1 for 

 the winter (Lam, 1968). Data were unavailable for 

 S. sagax, but like E. ringens it is a clupeoid, and the 

 reproductive seasons of the two species are similar 

 (Muck et al., 1987; Pauly and Soriano, 1987), there- 

 fore S. sagax was assigned the same seasonal values 

 as E. ringens. A value of 1.244 kcalg 1 was calcu- 

 lated for T. symmetricus by using proximate compo- 

 sition values from the related T. trachurus (Sidwell, 

 1981). Similarly, a value of 1.158 kcalg 1 for M. gayi 

 was based on equivalent data from M. productus 

 (Sidwell, 1981). For Loligo gahi, 0.968 kcal-g -1 was 

 obtained from proximate composition data in Croxall 

 and Prince (1982). No published values were avail- 

 able for Dosidicas gigas, so a mean ommastrephid 



