L T.Hztic St( Wit Vins C Pr4la4s Jsr , 5. T 3-1 .T1) . 6114 , SPECIES RE

Fr DC 02 cm-4. Ant-60,1cl i'arks &RC 1.44r•;.-, L t.hztic St( wit

viNs C Pr4La4s j s r , 5. t 3-1 .t1) . 6114 , -1

SPECIES RELATIONSHIPS AMONG FISHES OF THE GENUS1 GILA IN THE UPPER COLORADO RIVER DRAINAGE 2 2 3 Gerald R. Smith , Robert Rush Miller , and W. Daniel Sable

Taxonomic treatment of the chubs of the Gila hump, and coded continuous morphometric variables robusta complex (family Cyprinidae) that inhabit (23.1 15.8, etc.) into discontinuous states ("A", the Colorado River basin and certain rivers in "B", etc.), thus losing much of the resolving north-western Mexico (Figure 1) is problematical power of the measurements. The clustering illus- and requires study of the entire group for final trated by them shows only the level at which spec- resolution. It is intended to do this, but in this imens join clusters; it does not show intermedi- paper we present data demonstrating that there are acy, and it cannot display the morphological re- three species of this complex in the Colorado lationship of a specimen between other specimens. River and its major tributaries, from the Grand Thus there is no way to evaluate the supposed Canyon region upstream. These chubs are the most "intergrades" cited by Holden and Stalnaker: in extreme members of the genus in their specializa- their paper the "intergrades" may appear as a tion for life in the unique big-river habitat of result of the method of coding data. They were the Colorado River. unable to determine whether the so-called intergrades were part of one variable complex or two Our objectives are (1) to determine the valid- incompletely separated species. In contrast, we ity and relationships of the roundtail, bonytail, find that there are very few intermediates, in and humpback chubs in the big-river habitat; the Green River at least, and that most specimens, (2) to determine their relationships to other including the "intergrades" shown by Holden and populations of Gila in the middle and upper Stalnaker (1970: Figure 4), can be assigned to Colorado River basin; and (3) to find characters one of the two species cypha or elegans. useful for identifying young and adult Gila robusta Baird and Girard, G. elegans Baird and Our analysis consists of two parts. First, Girard, and G. cypha Miller, and possible hybrids. principal components analysis of 34 meristic and We do not at this time treat the complex repre- morphometric characters, not including a direct sentatives of Gila in the Gila River basin (see measure of the nuchal hump, is used to establish Rinne, 1976), except to note that G. intermedia clear evidence of morphological segregation of (Girard) may be a separate evolutionary line, the large-river specimens into three distinct secondarily related to G. robusta. clusters. It is important to establish that the segregation is not based on the nuchal hump or Misinterpretation of the patterns of variation other subjective characters, but on a broad range in the nuchal hump and lack of decisive, objective of body proportions, fin-ray counts, and verte- evidence has been the main cause of disagreement bral numbers, presumably with a broad genetic among those studying the G. robusta complex. The basis. Second, after establishing the existence nuchal convexity has never been measured precise- of three separate populations, with little or no ly or quantitatively related to other characters. overlap, at least within the unmodified, large- For example, Holden and Stalnaker (1970) used river habitat, individual discriminating key char- subjective code values for the form of the nuchal acters are developed for identification of these fishes in the laboratory and field. Such charac- 1 ters are critically needed to end the confusion Completion of this paper was delayed by clo- that has clouded attempts to formulate a sound sure of upper Green River to scientists other than management policy for these endangered or threa- federally funded personnel during the five-year tened fishes. The population analysis and the period 1963-1967. Considerable aid was received development of key characters were conducted from a number of organizations and individuals. separately by us in order to avoid circular taxo- Included are research grants to R. R. Miller from nomic logic and to enable the two systems to serve the Horace H. Rackham School of Graduate Studies as useful tests of one another. (1950), the National Science Foundation (G-15914, 24129, 24465, GB-3271, 4854, 6272X), and the Na- Principal components analysis, a multivariate tional Park Service (1975), with cooperation from statistical method enabling taxonomic use of com- the various western states for collecting permits. bined (correlated) information from many charac- Frances H. Miller recorded and calculated data, ters, was applied to 34 characters of 140 speci- helped prepare the distribution map, and criti- mens. These characters are: dorsal, anal, pec- cized drafts of the manuscript. Jeffrey N. toral, and pelvic fin-rays (rays from both paired Taylor prepared the original plots for Figure 1. fins recorded); standard length, head length, eye Loans of specimens were received from the U.S. diameter, snout length, preanal length, head National Museum (E. A. Lachner, S. H. Weitzman, depth through eye, head depth at occiput, inter- W. R. Taylor, Susan Karnella), Arizona State Uni- orbital width, occiput to tip of snout, dorsal- versity (W. L. Minckley), Bell Museum of Natural fin basal length, anal-fin basal length, predorsal History (Samuel Eddy, H. B. Tordoff, C. W. Huver), length, pectoral-fin length, pelvic-fin length, and the Museum of Northern Arizona (S. E. Caroth- upper jaw length, mouth width, body depth over ers). Mark J. Orsen (of the Museum of Zoology) pelvic insertion, caudal-peduncle depth, anal ' and Karna M. Steelquist (of the Museum of Paleon- origin to caudal base; number of vertebrae; pharyn- tology) prepared the drawings and other illustra- geal arches, total length, width, length of tions, except Figure 10, taken by Bruce J. Turner. anterior limb, and length of posterior limb, of Abbreviations are: UMMZ (University of Michigan both left and right arches (eight characters). Museum of Zoology) and USNM (U.S. National Museum This analysis discriminated the populations in of Natural History). question (Figures 2-5) showing the major trends 2 in variation, the characters dominant in the Museum of Zoology, University of Michigan, trends, and the characteristics of typical as well Ann Arbor, MI 48109. as intermediate individuals. 3 Wisconsin State University, Whitewater, A graph of 140 individuals plotted according to Wisc. 53190. their scores on principal components I and II, 613 121 117 113

25 0 25 50 100 200

SCALE OF MILES 50 0 50 100 200 300

SCALE OF KILOMETERS

1 - Gila River • Gila reboto 2 - Little Colorado River O Gila cyplso 3 - Colorado River O Gila elegans I • Son Juan River 5 - Green River

117 113 109 105

FIGURE 1. Distribution of Gila robusta, Gila elegans, and Gila cypha. See Rii-i-fteRAPme (1976) for distribution of allied forms in the Gila basin.

614 •

76 %

elegans U. 9% 00 00 o o 00 0 o 0 0 0 0

0 0 q0 cypha 0 0 0 0 0 0 /

0 o 0 0 O 0 ® 0 CE1 En 0 0 00 O DOD 0 ® 0 0 O DD 0 _04* DO 0 00 0 • • 0 0 0 0 0 • • O 00 ® • e 8 •• ® *seminude ® ••• • .• .:• ----- 0 • % IB • *io • • g • • • - • • robusta ------..,....„...... ,...... ••

FIGURE 2. Principal components analysis, components I and II, of four forms of Gila. G. robusta robusta, G. elegans, and G. Cypha are discriminated by the combination of projections shown in Figure 2 and 3. G. r. seminuda is scat- tered through the other clusters. Percentages indicate the proportion of the total variation accounted for by each component.

TABLE 1. Development of nuchal hump in Colorado River chubs, Genus clia.a

Species Ratio (range) S.L.range (mm)

cypha (31) 6-13 206-328 elegans (20) 15-29 245-412 r. robusta (17) 28-207 211-309 r. seminuda (4) 31-121 175-247

a — Expressed as ratio: Pi rn I '-'2 distance Depth frontal depression Figures in parentheses indicate number of specimens measured.

615 the Colorado and Green rivers. This conclusion for the necessary stability to persist. cannot be confidently applied to certain populations that we have available but have not fully In a period of controversy following the poi- studied from Lake Powell, however. Furthermore, soning of the upper Green River in 1962 (Miller populations of Gila robusta from certain tribu- 1963b), erroneous statements about the classifi- h/ taries ,ere elegans and cypha are absent parallel cation, biology, and distribution of Colorado these species in many characteristics. We choose River chubs appeared which have been destructive a taxonomic treatment of this situation that em- to the understanding and management of these phasizes the specific distinction of the three fishes, and require correction. For example, populations in the big-river habitat, but admit Stroud (1963:7) irresponsibly claimed that the the possibility that the species isolating mecha- construction of Flaming Gorge Reservoir resulted nisms may break down under disturbed (reservoir) in making available "large numbers" of humpback conditions and that the populations in tributary chubs, all of which were supposedly males. The streams may not be completely independent or iso- agency (U.S. Bureau of Sport Fisheries and Wild- lated from the three central forms. We refer to life) that "confirmed the presence and ready the entire complex as the Gila robusta super- availability of numerous humpback chubs in [the species, including robusta, cypha, and elegans Green River below Flaming Gorge Dam] . . ." as defined above, and including subspecies or subsequently reported (Vanicek 1967; Vanicek et races of robusta, seminuda of the Virgin River, a al. 1970; Holden and Stalnaker 1975) the absence similar population in the San Juan River, and a of Gila cypha in the area indicated and great number of isolated populations in northern and rarity of this species below the mouth of the western Mexico. The complex in the Gila River Yampa River; only three specimens were taken in basin includes Gila intermedia and forms inter- 1963, and none during 1964-1966, in the Green mediate between that species and robusta, i.e., River in Colorado and Utah (Kramer 1967:Table 2; G. r. "grahami" (cf. Rinne, 1976). Gila inter- Vanicek et al. 1970:Table 4). media is inferred to be a part of the robusta superspecies on the basis of shared characters Vanicek and Kramer (1969:194-195) stated, and possible gene exchange through "grahami", but ". . . criteria are not available for distinguish- may be a sister phyletic line by its origins, as ing between young fish of the two morphological indicated by its similarity to several of the variants" (i.e., between what were then called large-scaled, small-finned subgenera of Gila (cf. Gila r. robusta and G. r. elegans). "Consequent- Miller, 1945). ly, specimens shorter than 200 mm total length were combined in the present study under the In spite of the uncertainty regarding peripher- general taxon, Colorado chub." Collections of al populations, it is clear that robusta (s.s.), Gila robusta and G. elegans at The University of elegans, and cypha have diverged in a number of Michigan, containing young as small as 22 to 40 mm characters, and can be discriminated on the basis S.L., were identified as early as 1926; by 1968, of counts of fin rays, vertebrae, and gill rakers, humpback chubs as small as 43 mm S.L. had been the relative depth and length of the caudal pe- determined. By the criteria reported here, we duncle, snout shape, fin position, and form of the have identified Gila robusta robusta (UMMZ 162818) nuchal hump. The multivariate analysis shows that to 20 mm, Gila cypha (UMMZ 182415) to 54 mm, and cypha is intermediate between robusta and elegans Gila elegans (UMMZ 162846) to 22 mm in total in general, but is extreme in several respects. length. Juveniles of three species are shown in This suggests the possibility that cypha and Figure 10. Holden and Stalnaker (1975) (citing elegans were derived from robusta by separate Minckley and Deacon, 1968, as authority) reported speciation events. Gila robusta shows the most Colorado squawfish (Ptychocheilus lucius Girard) primitive characters of the three, and the sepa- in the Grand Canyon, but the two specimens actu- rate ways in which elegans and cypha are extreme ally came from between Glen Canyon Dam and Lees suggest that neither is likely to be ancestral to Ferry, well above Grand Canyon. The only valid either of the other two. record (based on preserved material) known to us of this species from the Colorado River in Grand The lack of coexistence of the three species Canyon is represented by an adult (about 320 mm in smaller tributaries is in accordance with the S.L.) caught in 1975 by an unknown fisherman at expected relationship between diversity and spatial the mouth of Havasu Creek (ASU 7087). However, heterogeneity. The species seem to be omnivorous squawfish formerly moved up the Little Colorado carnivores with specializations related to habitat River in Grand Canyon to the base of Grand Falls --ch4Qters associated with food processing (jaws, (Miller, 1963a:1, ftn.1). teeth) are similar, except that robusta has fewer gill rakers and elegans more (Table 2). In small In our distribution map of the Gila robusta tributaries, such as the Virgin and possibly the complex (Figure 1), the record for G. elegans in San Juan, a single species with intermediate morph- the Little Colorado River (at base of Grand Falls) ology seems to be selected for; in the larger is based in part on the statement referred to tributary systems of the Gila River, a bewilder- above, in part on our conclusion as to the true ing mosaic of generalized forms occur (Rinne, 1976), type locality for Gila elegans (and Gila robusta), where in former times robusta and elegans lived and in part on the absence of the bonytail from in the main channel. the Little Colorado River above Grand Falls. The holytype (USNM 251), as well as the three syntypes of Gila robusta (USNM 246), were said to have The close apparent control of habitat size and come from the "Zuni River, New Mexico" (Baird and diversity over morphology and species diversity Girard, 1853, 1854; Girard, 1858:286-287). How- suggests that if the habitat could be experimen- ever, at the time of collection during the summer tally changed, the populations should show pre- (rainy season) of 1852, Zuni River was described dictable responses. Rather unfortunately, the ". . . as a mere rivulet, and not entitled to the destruction of main-river habitat by Glen Canyon name of river; in most parts of our country it Dam has created such an experiment. Early indi- would not be dignified with that of creek" cations are that the two more specialized species (Sitgreaves, 1854:5). This is hardly the habitat will not exist in the Lake Powell environment. of Gila elegans and, moreover, that species is The three species seem to be breaking down locally unknown from the Little Colorado River basin (to by hybridization. We predict that a single- which Zuni River is tributary in floods) above species of mixed origins and generalized charac- Grand Falls, an impassible barrier 56 meters high teristics will appear. Given long enough, some (Dryer, 1965). Furthermore, careful examination re-oriented diversification would develop, but of the channel of Zuni River in New Mexico con- the reservoir is apparently silting in too rapidly 622 vinced us that (at least in recent centuries) •

this river did not provide a suitable habitat for ESCHMEYER, W. N., and S. G. POSS. 1977. Review the bonytail (W. L. Minckley, who has also stud- of the scorpionfish genus Maxillicosta (Pisces: ied the channel, concurs--pers. comm. 1976). Scorpaenidae), with a description of three new Study of Sitgreaves report and map shows that his species from the Australian-New Zealand region. last station in the basin of the Little Colorado Bull. Mar. Sci. 26(4):433-449. River (No. 13) was at Grand Falls. We thus assume GIRARD, C. 1858. Fishes. In: General report that this is the correct type locality for Gila upon the zoology of the several Pacific rail- elegans and G. robusta (of which the third species road routes. U.S. Pac. R.R. Expd. and Surv. taken, G. gracilis, is a synonym). Incidentally, 10(4):i-xiv, 1-400. the genus Gila was named for the river of that HOLDEN, P. B., and C. B. STALNAKER. 1970. Sys- name on the mistaken idea that the Little Colorado tematic studies of the cyprinid genus Gila, in River was tributary to Salt River of the Gila the upper Colorado River basin. Copeia 1970 River basin. (3):409-420. HOLDEN, P. B., and C. B. STALNAKER. 1975. Dis- The perpetuation and management of robusta, tribution and abundance of mainstream fishes elegans, and cypha depends on the maintenance of of the middle and upper Colorado River basins, large sections of unspoiled, natural, main-channel 1967-1973. Trans. Amer. Fish. Soc. 104(2): habitat. Dams threaten these species primarily by 217-231. destroying the swift-water habitat above the ob- KRAMER, R. H. 1967. Introduction, pp. 1-9. In: struction and drastically decreasing the tempera- Green River fishes and invertebrates. Spec. ture and turbidity below. Introduced exotic com- Repot., U.S. Bur. Sport Fish. and Wildlife, petitors and predators (on young) are a threat, Utah Coop. Fish. Unit, Logan, Utah. especially in these modified areas. The most im- MILLER, R. R. 1945. A new cyprinid fish from portant current need is research on the habitat, southern Arizona, and Sonora, Mexico, with food, reproduction, recruitment, and hydrodynamics the description of a new subgenus of Gila and a and energetics of these forms of Gila, both for review of related species. Copeia 1945(2):104- the purpose of sound management and for the reason 110. that such information is one of the valuable bene- MILLER, R. R. 1963a. Distribution, variation, fits that man can derive from these unique species. and ecology of Lepidomeda vittata, a rare cyprinid endemic to eastern Arizona. Copeia Key to Species of Gila in the Main-channel Habitat 1963(1):1-5. of the Upper Colorado River Basin* MILLER, R. R. 1963b. Is our native underwater life worth saving? Nat. Parks Mag. 37(188): la. No nuchal hump, frontals nearly straight 4-9. • (Figures 8,10 above), dorsal and anal rays usually RINNE, JOHN N. 1976. Cyprinid fishes of the 9, distance from anal origin (base of first ray) genus Gila from the lower Colorado River basin. to caudal base (bases of rays, internally) much Wassman J. Biol. 34(1):65-107. less than distance from anal origin to opercle SITGREAVES, CAPT. L. 1854. Report of an expe- ...... robusta. dition down the Zuni and Colorado rivers. U. S. Senate, 33rd Cong., 1st Sess., Exec. lb. Nuchal hump present in specimens over 150 mm Doc., Washington, 198 pp. (Figures 6,9), frontals concave above eyes in STROUD, R. H. 1963. Green River humpbacks. lateral profile (Figures 6,9,10 middle and below), Sport Fish. Inst. Bull. 141:7. dorsal rays 9 or 10, anal rays 10 or 11, distance VANICEK, C. D. 1967. Ecological studies of from anal origin to caudal base about equal to or native Green River fishes below Flaming Gorge greater than distance from anal origin to opercle Dam, 1964-1966. Ph.D. Thesis, Utah State 2 Univ. Issued as Spec. Rept., U.S. Bur. Sport Fish. and Wildlife, Utah Coop. Fish. Unit, 2a. Nuchal hump abrupt (Figure 6), dorsal rays Logan, 124 pp. usually 9, anal rays usually 10, in young less VANICEK, C. D., and R. H. KRAMER. 1969. Life than 150 mm the eye diameter less than 2/3 caudal history of the Colorado squawfish, Ptycho- peduncle depth, snout overhangs upper lip (Figure cheilus lucius, and the Colorado Chub, Gila 10 middle) ...... cypha. robusta, in the Green River in Dinosaur National Monument, 1964-1966. Trans. Amer. 2b. Nuchal hump not abrupt (Figure 9), dorsal Fish. Soc. 98(2):193-208. rays usually 10, anal rays 10 or 11, in young less VANICEK, C. D., R. H. KRAMER, and D. R. FRANKLIN. than 150 mm the eye diameter greater than 2/3 1970. Distribution of Green River fishes in caudal peduncle depth, snout not overhanging Utah and Colorado following closure of Flaming upper lip (Figure 10 below) ...... elegans. Gorge Dam. Southwest. Nat. 14(3):297-315.

*Supplemental characters in Tables 1-3 aid identification, especially for difficult specimens and possible hybrids. Hybrids tend to have several characteristics distinctive of each parent and several intermediate. Specimens may be sent to The University of Michigan Museum of Zoology for identification.

LITERATURE CITED

BAIRD, S. F., and CHARLES GIRARD. 1853. Descrip- tions of some new fishes from the River Zuni. Proc. Acad. Nat. Sci. Phila. 6:368-369. BAIRD, S. F., and CHARLES GIRARD. 1854. Fishes, pp. 148-152, pls. 1-3. In Sitgreaves, Capt. L. Report of an expedition down the Zuni and Colorado rivers. U.S. Senate, 33rd Cong., 1st Sess., Exec. Doc., Washington. DRYER, I. 1965. Grand Falls of the Little Colo- rado. Nat. Parks Mag. 39(212):10-12.

623