7829

• UCREFRP Catalog Number: 7829
• Author: McElroy, D. M. and M. E. Douglas
• Title: Patterns of Morphological Variation among Endangered Populations of
• USFW Year: 1995
• Copyright Material: YES

642
0.55
U 0.4
a
COPEIA, 1995, NO. 3
0
0 0
0 0 00 0
° 0° °
• • •• • 0 0 0
•o ?• 'Op 8000 °
M 0 00 o °to °do % o
Mb A A o °X 180
omOYd am(s)
• 0 a °°°QL0
%
•
r
0.25 4-
0.02
0.15
PC 11
a
4
2
U 0
-2
0.28 -4

?A
0
v o?o? o a •
o • e• as :0
0 A

?
O• 641 *A
•
,
°
?'j
3
o?
?
®? •• •
• O

C6 0
_
A
o
V, o
y
•?
tt
o ?'?S ? y

o • %
-8 -6 -4 -2 0
CV 1
U
C, D
C, D
Specimen
Fig. 6. Canonical variates analysis of morpholog-
ical variation within and between Gila robusta and G.
b cypha. CV I and II represent the first two shape factors
extracted from this analysis. There is a clear sepa-
ration of the following: (1) a G. robusta group (closed
circles) composed of specimens from Black Rocks,
Westwater and Debeque canyons, Rifle, and the Yam-
pa River; (2) a G. cypha group (open circles) repre-
senting samples from Black Rocks, Westwater and
Grand canyons, and the Yampa River; and (3) a mixed
group of G. robusta (open triangles) and G. cypha (closed
triangles) from Cataract and Desolation canyons.
Fig. 5. Principal component (a) and canonical var-
iates (b) analyses of morphological divergence be-
tween specimens of Gila robusta and G. cypha. In (a),
axes represent the first two shape factors derived from
the analysis. PCA provides evidence of structure be-
tween G. robusta (closed circles) and G. cypha (open
circles), but within-species point clouds overlap con-
siderably. In (b), the sample of G. robusta (specimens
1-215) clearly differs from that of G. cypha (specimens
216-363). Samples of both species taken from Cata-
ract and Desolation canyons (indicated by labeled bars)
are, on average, more intermediate (scores closer to
zero) than are other samples.
ciated with relative size and positions of the
dorsal and anal fins (ODo-OA, OA-vIA, OA-
IA) were characterized by having the largest
univariate F values. All specimens were cor-
rectly classified with DFA.
Geographic patterns of variation.-In addition to
clear separation of species, analyses of variation
at the generic level indicated the presence of a
strong locality effect on relationships among
several populations. The separation of groups
(e.g., populations) through canonical variates
analysis was good; 55 of 56 univariate character
ANOVAs displayed a significant population ef-
fect (P < 0.0009), and multivariate tests were
highly significant (ex. Wilk's Lambda = 0.0002,
F672,3510.7 = 2.314, P < 0.0001). As before, char-
acters describing fin placement (OA-vIA, ODo-
OA, OA-IDo) showed the largest univariate F
values. Nine of 12 canonical roots contained a
significant component of among-group varia-
tion. CVI and II produced a clear separation of
three groups: (1) a G. cypha group consisting of
individuals from Black Rocks, Westwater and
Grand canyons, and the Yampa River; (2) a G.
robusta group consisting of fish from Black
Rocks, Westwater and Debeque canyons, Rifle,
and the Yampa River; and (3) a mixed group
of G. robusta and G. cypha individuals from Des-
olation and Cataract canyons (Fig. 6). Greater
than 92% (336 of 363) of all specimens were
correctly classified using DFA. In addition, de-
spite an overall classification error rate of 0.075
(27 of 363 misclassified), the majority (89%) of
misclassifications occurred within rather than
among species (Table 2). Three specimens of
G. cypha were assigned incorrectly to G. robusta
populations; no individuals of G. robusta were
classified as G. cypha.
Cluster analysis of generalized canonical dis-
tances among populations produced two pri-
mary, and somewhat contradictory, results. First,
conspecific populations tended to cluster to-
gether to the exclusion of heterospecifics (Fig.
7); samples of G. robusta and G. cypha generally