8235 - Laserfiche WebLink

Home Browse Search

DISCRIMINATION OF COLORADO PLATEAU GILA SPP. 169 than G. elegans. It is also worth noting that depth of frontal depression (a direct measure of nuchal concavity) was not a major component of species segregation. However, a more indirect measure (head depth at occiput) was. Gila robusta and G. cypha have been a focus for the majority of morphometric research accom- plished on this group to date. Douglas et al. (1989) used qualitative characteristics to separate these species in the Yampa River. Kaeding et al. (1990) compared field identification at Black Rocks, Col- orado, with taxonomic assignment based on a prin- cipal component analysis (PCA). In the former, the complex relationships between anal fin and caudal peduncle, eye and jaw, as well as snout, occiput, and breast-nuchal scalation, were deemed impor- tant; in the latter, the most important characters were depth of nuchal depression and caudal pe- duncle depth. McElroy and Douglas (1995) ex- amined population-level differentiation of these species from eight sites in the upper Colorado Riv- er basin, and McElroy et al. (1997) differentiated the two species on four morphometric characters (eye diameter, caudal peduncle, orbit-jaw relation, and skull depression-nuchal hump). In this study, morphometric characters (Table I) best segregated the two species, and depth of fron- tal depression was an important character, which was also the case for Kaeding et al. (1990) and McElroy et al. (1997). A similar level of discrim- ination is maintained in the field subset, even though the former character was removed (albeit with the addition of two additional morphometric characters). However, other morphometric char- acters listed in Table 1 have not been previously noted as discriminatory. For example, two of four field characters (anal fin base and head depth at eye) relate more to qualitative characters depicted in Douglas et al. (1989: Figure 2) than to any oth- ers. Two remaining field characters (dorsal fin base and pectoral insertion to pelvic insertion) have not been previously identified as important. Their des- ignation supports Douglas et al. (1989), who ar- gued for application of new approaches and new characters in resolving this complex problem. Smith et al. (1979) evaluated phenotypic rela- tionships among all three study species using PCA and noted that characters best contributing to their separation were dorsal and anal rays, gill raker and vertebral number, and depth and length of caudal peduncle. Those meristic characters indicated by Smith et al. (1979) also discriminated well in our analyses (Table 1). In addition, caudal peduncle depth was an influential morphometric character, as were several others not mentioned by these au- thors. Discriminating field characters pertained to the peduncle, its relation to the anal fin, the snout, its relation to the occiput, and the dorsal fin base. All three data sets discriminated greater than 95%. It is clear from this study that a variety of char- acters differ significantly among these species, and various combinations of characters provide excel- lent discrimination. The high kappa value (P < 0.00001) and failure of the jackknife to improve the original segregation underscore this point. Fur- thermore, misclassified individuals, both within each species-pair and among all three species, were rarely shared among the three data subsets, again suggesting that a variety of characters ade- quately discriminate. We thus reject the hypothesis that the three species are phenotypically indistin- guishable. It is also clear that G. elegans is the most dif- ficult of the three to assign to group. Of four mis- identified individuals in the discriminant analysis with field data, three were G. elegans (Figure 2). Similarly, when sample sizes were equilibrated among species and within-group covariances were used in the analysis, two (of three) misidentified individuals were G. elegans. Assignment of Juveniles and Putative Hybrids to Species Overall, assignment of juveniles to species seemed reasonable, with 88% correctly allocated. However, this percentage is misleading. The pre- ponderance of juveniles were G. robusta (N = 106; 75%), and these were correctly assigned at greater than 97%. Correct assignment of juveniles to G. cypha and G. elegans were only 60% and 66%, respectively. Furthermore, if one considers that several correct assignments of juveniles to these species were actually borderline with respect to score, then the number of unambiguous assign- ments is even weaker. This suggests that although field characters (Table 1) may be satisfactory for designating juvenile G. robusta to species, they appear marginal at best in assigning juvenile G. cypha and G. elegans. Quite possibly, the occiput and nuchal hump have not as yet developed ade- quately enough to serve as the diagnostic tool these structures provide in adults (Table 1). Given these mixed results, we cannot reject our second hy- pothesis, that adult characters also segregate ju- veniles. However, this does not mean that juveniles of G. cypha and G. elegans cannot be differentiated phenotypically. Composites of several different quantitative characters may instead be useful;