9336 - Laserfiche WebLink

Home Browse Search

392 COPEIA, 2001, NO.2 energy matrix. Relative warps are principal com- ponents of this space, and describe the major trends in shape variation within the overall sam- ple (Rohlf, 1993). Various forms of RWA can be obtained by varying the exponential weight for the bending energy metric (0:) between zero and one. When 0: > 0, geometrically small-scale variation is given less weight than large scale var- iation. This has the effect of reducing the weight given to regions having more landmarks relative to that given to regions having fewer (and hence more widely spaced) landmarks. We used 0: = 1 because we were intereste i in mor- phometric differences at a larger scale. In ad- dition, we were looking for shape variation in regions other than the nuchal hump, which comprised 40% (10/25) of our landmarks (Fig. 1). It also corresponds to the relative warp anal- ysis of Bookstein (l989b, 1991; see Bookstein, 1996:146-147). As noted by Rohlf et al. (1996), the value of 0: has no effect on most of the anal- yses employed herein because it merely scales the partial warps. The results of most multivar- iate analyses (e.g., MANOVA, CVA, for exam- ple) are scale free. For our analysis, mean shape-coordinate forms were calculated for all 13 samples, with the grand mean as the tangent configuration. We also compared patterns of morphological shape variation in G. robusta and G. CYPha by de- riving taxonomic distances among the five sym- patric populations only, based upon population means for relative warp axes one and two, de- rived from GLS-superimposed specimens. Both matrices were compared against a matrix of geo- graphic distances derived from latitude-longi- tude values and against one another using two- way Mantel tests. RESULTS Body size and shape variation in Gila CYPha.-Cen- troid size varied significantly among popula- tions (Fs.142 = 7.33; P < 0.0001) with Little Col- orado River specimens significantly larger than all other groups (Student-Neuman-Keuls test). All shape coordinate pairs showed significant differences among populations (all MANOVA, P < 0.0001), and the overall MANOVA was also significant (Wilks' A = 0.002; X2240 = 743.47; P < 0.0001). Variation can be most clearly seen by examining shape coordinate plots among samples (Appendix 1). Three aspects are em- phasized: shape of the nuchal hump; relative size of the head; and dimensions of the caudal peduncle. The nuchal hump is relatively more pronounced and anteriorly projected in Little Colorado River G. CYPha (Appendix ID), where- CV2 @ 2 G e 2 CV1 8 @ @ Fig. 2. Shape variation among six Gila typha pop- ulations based on the first two variates derived from a canonical analysis of shape coordinates. Population abbreviations follow Table 1. Circles represent 95% confidence intervals for the group mean. as specimens from Cataract Canyon (Appendix IB) have a relatively slight hump. The position of the three landmarks bounding the opercu- lum (LMs 16, 18 and 19; Fig. 1) indicate that specimens from Cataract Canyon have relatively larger heads than do those from the Little Col- orado River, whereas all other populations are intermediate. Finally, there is significant varia- tion in lengths and depths of the caudal pedun- cle, with specimens from Cataract Canyon hav- ing peduncles that are relatively shorter in length but that taper in depth from anterior to posterior, particularly when compared to those from Black Rocks Canyon (Appendix lA), West- water Canyon (Appendix IE), or Yampa River (Appendix IF). Desolation Canyon individuals (Appendix lC) have peduncles that are uni- formly deeper and non tapering than all others. Other aspects of shape variation are found in the distance from pectoral to pelvic fins (LMs 2-20), which is reduced in the Little Colorado River population. Fish from this population also have the longest anal fin base, which is reduced in specimens from Cataract Canyon. Also, Des- olation Canyon individuals have shallowest body depths. A canonical variate analysis of shape coordi- nates resulted in three statistically significant variates (of five), accounting for 95.9% of the among-population variation (P < 0.0001; Fig. 2). Again, Little Colorado specimens were well separated from the other five populations. All Mahalanobis D2 distances were also significant (P < 0.0001) and demonstrated a high degree of among-sample variation. Classification results were significantly different from chance, with