Laserfiche WebLink
640 <br />2.25 <br />0 <br />U <br />-2.25 <br />-4.5 <br />COPEIA, 1995, NO. 3 <br /> ° <br />e <br /> <br />° ° ° <br />• o <br />• •O 93 <br />oftf <br />e • ° Ly1h <br />° •q? <br />B <br /> °0 o <br />A& <br /> ° • a <br />o <br /> ?I • <br /> 0 <br />• <br />i <br /> • t <br />• <br />s'~=•• <br /> ' s.. <br />-4.5 -2.25 0 2.25 4.5 <br />CV I <br />p 0 <br />U -2 <br />-4 <br />-6 <br />-g <br />O ° <br />a <br />O •?®°? ° ° All A <br />• ••? •s b • <br />d ° • • • • <br />m a <br />ant, 0 <br />m6 6, <br />a <br />an e <br />e <br />a <br />-4 -2 <br />CV I <br />a <br />b <br />Fig. 3. Canonical variates analyses of population <br />divergence in (a) Gila robusta and (b) G. cypha. Axes <br />represent the first two shape factors extracted from <br />the analyses. Significant among-group heterogeneity <br />exists on CVI and CVII in both species, and all pop- <br />ulations differ significantly in CV space when higher <br />dimensions are considered. Group labels are as fol- <br />lows: Black Rocks (open squares), Cataract Canyon <br />(closed circles), Desolation Canyon (open triangles), <br />Debeque Canyon (closed triangles), Grand Canyon <br />(hatched squares), Rifle (open circles), Westwater <br />Canyon (closed diamonds), Yampa River (open dia- <br />monds). <br />using a Mantel test (Mantel, 1967); here, the <br />normalized Mantel statistic computed from the <br />original data was compared to a sample distri- <br />bution based on 1000 permutations of the geo- <br />graphic matrix using NTSYS-pc (Rohlf, 1992). <br />Statistical analyses.-Patterns of morphological <br />variation in Gila were assessed by applying the <br />above methods to four statistical models, rep- <br />resenting three levels of biological organiza- <br />tion. First, the degree of population divergence <br />within G. robusta and G. cypha was examined in <br />separate analyses by grouping fish according to <br />their population of origin (n,°uPs = 7 for G. ro- <br />busta, ng.°uPs = 6 for G. cypha). Next, morpho- <br />logical distinctiveness at the species level was <br />evaluated by categorizing individuals as G. ro- <br />busta or G. cypha, independent of their popu- <br />lation of origin. Finally, geographic patterns of <br />morphological variation at the generic level were <br />investigated by combining the first two analyses <br />into a single model; that is, individuals were <br />grouped by species and population (n,,-,,, = 13) <br />to simultaneously evaluate the relationship <br />among sympatric and allopatric populations. <br />In each model, field identifications were used <br />to group individuals by species. Although a more <br />robust test of morphological variation would <br />rely on genetic data or some other independent <br />character to determine affinity of specimens, <br />our approach is valid as a test of the nature and <br />consistency of qualitative field identifications and <br />the relative distinctiveness of species/popula- <br />tions among localities. <br />RESULTS <br />Variation in G. robusta.-Although principal <br />components analysis implied limited population <br />structure in G. robusta, canonical variates anal- <br />ysis revealed significant among-group diver- <br />gence in shape. Individuals differed primarily <br />in size; the first principal component of the vari- <br />ance-covariance matrix (PCI) was characterized <br />by consistently high, positive loadings across all <br />characters and explained 82.2% of the total <br />variation. By contrast, PCII-IV combined ex- <br />plained only 47.4% of the residual variation and <br />failed to clearly separate any populations. At <br />least 16 latent roots (excluding PC I) contained <br />significant information (Reyment, 1992). <br />Despite the high degree of overall variability <br />evident from PCA, populations could be clearly <br />differentiated with CVA. Forty-nine of 56 size- <br />corrected characters displayed significant uni- <br />variate differences among groups based on con- <br />servative Bonferroni-corrected criteria (P < <br />0.0009); multivariate tests of among-group dif- <br />ferentiation also were highly significant (Wilk's <br />Lambda = 0.0032, F„6,922.6 = 4.4542, P < <br />0.0001). Characters associated with position of <br />the pectoral and pelvic fins and body depth (vPe- <br />OPe, VPd-Dpd, OPl-ODo) displayed the largest <br />univariate F values. Each of the six canonical <br />roots carried significant among-group structure <br />(P < 0.05). The first canonical vector (CVI) <br />separated the Desolation Canyon and, to a less- <br />er extent, Cataract Canyon populations from <br />all others (Fig. 3a). Specimens of G, robusta could <br />be assigned to groups with a high degree of <br />confidence using DFA. The overall classifica-