Pacific Science (1996), vol. 50, no. 2: 184-193 © 1996 by University of Hawai'i Press. All rights reserved Morphological Variation in Feeding Traits of Native Hawaiian Stream Fishes 1 MICHAEL H. Kmo2 ABSTRACT: The five native species of amphidromous gobioid fishes inhabit­ ing Hawaiian streams were compared for dentition, gut length to body length ratios, intestinal convolution, gill raker morphology, position of mouth, and diet. Based on morphological comparisons, three manipulative modes of feed­ ing are indicated, as follows: picking-biting, rock scraping, and sediment for­ aging. Comparisons indicated a surprising predominance of algae in the diet of all species despite various degrees of morphological specialization for their use. Avoidance of competition for algae was therefore suggested as a potential fac­ tor influencing species interactions and community organization. Differential preference among native gobioids for stream invertebrates may also provide mitigation for competitive interactions. Variation in food availability in the benthic landscape of Hawaiian streams, possibly regulated by stream flow and periodic disturbance, is hypothesized as being an important determinant of fish community structure. Human-induced alteration of factors that regulate food availability could therefore influence stability ofnative fish populations through disturbance of their food base. COMPARING SPECIES DIFFERENCES in morpho­ species that do not associate (Grant and logical traits for acquiring food and explor­ Schluter 1984). Of central ecological inter­ ing the relationship of these characters to est is the role of competition in influencing resource use is a fundamental step toward community organization, and morphological understanding trophic interactions and pat­ comparisons can provide a "firm base" for terns of community organization in fish. such inquiry (Maiorana 1978). In this study, These traits affect food selection and foraging I examine the validity and applicability of efficiency and establish a basis for predicting these tenets for amphidromous stream fishes food and habitat use among fish species native to the Hawaiian Islands. (Werner 1984). Dietary differences between Only five fish species in two families species should correlate with morphological (Gobiidae and Eleotridae) are native to the differences (Schoener 1965, Schluter 1982), steep-gradient mountainous streams of Ha­ and it is expected that species that coexist wai'i, where the native aquatic food base is should be less similar in morphology than largely limited to algae and immature Dip­ tera. It seems plausible that competition for this limited diversity of foods, spurred by the extreme isolation of the Hawaiian Archipel­ 1 This study was made possible through program sup­ ago, has played an evolutionary role in shap­ port to the Kaua'j Research Facility (University of ing species interactions (Kido, unpubl. data). Hawai'i, College of Tropical Agriculture and Human Resources) for environmental research in the Hawaiian Dietary studies on these gobioids, however, Islands from the Agricultural Research Setvice, U.S. are incomplete and very little of their basic Department of Agriculture and through a grant from the biology has been studied; thus virtually noth­ Division of Aquatic Resources, State of Hawai'i, De­ ing is known about morphological variation partment of Land and Natural Resources, Award No. LOIO-P0591O. Manuscript accepted 15 June 1995. between species ortheir connection to resource 2 Kaua'i Research Facility, University of Hawai'i, use. Published diet information is only avail­ 7370A Kuamo'o Road, Kapa'a, Hawai'i 96746. able for two ofthe gobies (Gobiidae). Awaous 184 Morphological Variation in Fish Feeding Traits-KlDo 185 guamensis (Valenciennes) is described as a MATERIALS AND METHODS generalist that utilizes both algae and inver­ tebrates (Kido et al. 1993), and Sicyopterus The study was conducted in the Wainiha stimpsoni (Gill) has been shown to special­ River, which drains the Alaka'i Swamp atop ize on diatoms and blue-green algae (Kido the ancient caldera of Mount Wai'ale'ale on 1996). These two species exhibit a distinct the island of Kaua'i. Discharging into the pattern in the partitioning of algal foods and ocean on the island's northern shore, Wai­ strong preference for the ubiquitous green niha River descends to sea level from an ele­ alga Cladophora sp. (Kido, unpubl. data). vation of 610 m, traveling a distance ofca. 21 No published dietary information is avail­ km. Mean discharge for 1992 measured by a able for Lentipes concolor Gill and Steno­ permanent U.S. Geological Survey gauge at 3 gobius hawaiiensis (Cuvier & Valenciennes) 300 m elevation was 3.5 m sec -1. Four fish (both Gobiidae) or the lone native eleotrid collection sites were used, at sea level (es­ (Eleotridae), Eleotris sandwicensis Vaillant & tuary), ca. 37 m elevation (site 2), ca. 116 m Sauvage. elevation (site 3), and in the steeply graded Population structure of gobioids among tributary, Maunahina, at ca. 190 m elevation Hawaiian Island streams also has been poorly (site 4). Stored specimens of S. stimpsoni and studied both spatially and temporally. Eleo­ A. guamensis, collected previously in sites tris sandwicensis and S. hawaiiensis overlap 2 and 3 (February 1992 to January 1993) in range and generally are confined to lower (Kido, unpubl. data), were utilized for mor­ elevations and estuaries (Maciolek 1981). phological comparisons. Lentipes concolor The remaining three gobies are less restricted was collected in Maunahina (site 4) and E. and range farther into mountainous reaches; sandwicensis from site 2 in August 1992, however little quantitative data on species whereas S. hawaiiensis was collected in Wai­ distribution are available. Coexistence and niha estuary in July 1993. positive association, which can be influenced Adult fishes were captured using hand­ by species densities, were found for A. gua­ nets in the estuary and with electrofish­ mensis and S. stimpsoni (Kido, unpubl. data). ing gear in the mountainous sites. Fishes Based on studies in three streams on three were anesthetized in the field with MS-222 Hawaiian islands, Kinzie (1988) determined (tricaine methanesulphonate), measured for that L. concolor did not co-occur with other standard length, weighed to the nearest 0.1 g, gobies because ofits longitudinal distribution and preserved in 10% buffered formalin. In pattern and that A. guamensis and S. stimp­ the laboratory, fishes were examined exter­ soni differed in microhabitat utilization pat­ nally for morphological features related to tern in those reaches of streams where they feeding (Lowe-McConnell 1978). The diges­ co-occurred. How are these patterns related tive system was dissected, examined for to resource use and guided by species differ­ arrangement of folds, and sketched using a ences in morphological feeding traits? dissecting microscope (Leica-Wild M37). All The following objectives were addressed in gut drawings were made viewing the organ this study: (1) to compare the morphological ventrally (Fukusho 1969). The pattern of feeding traits of the five native Hawaiian convolution was diagnosed and compared by stream fishes and determine the availability indicating positions of turning or bending of feeding modes; (2) to determine the degree (Fukusho 1969). The gut subsequently was to which these differences are reflected in re­ removed from esophagus to anus and stored source use; and (3) to assess the role of food in 10% buffered formalin. Digestive systems competition in structuring native stream fish were later unraveled and measured for total populations. This information is of broad length to calculate gut length to standard ecological interest but is also useful for de­ length ratios (Lowe-McConnell 1978). The veloping effective management strategies for gut was opened and the animal and plant declining native stream fish populations in portions of the contents were sorted, identi­ the Hawaiian Islands. fied to lowest possible taxonomic category 186 PACIFIC SCIENCE, Volume 50, April 1996 (as in Kido 1996), dried at 60°C for 48 hr, random from all size classes from stored and weighed to the nearest 0.0001 g. material. Mean standard length ofthese fishes Five morphological feeding traits were was 95.1 ± 2.91 and 74.9 ± 1.67 mm, respec­ compared in the study. For comparisons of tively. All species except E. sandwicensis bony internal features of the mouth, fishes had ventrot"t:rminal mouths with the lower were cleared and stained with alizarin red jaw closing posterior to the upper jaw. using an enzyme method (Taylor 1967) that Eleotris sandwicensis differed in having a ter­ removes flesh with minimal damage to bony minal mouth with a wide gape that opened structures. Gut length to standard length somewhat dorsally, the lower jaw closing ratios were normalized using a log (x + 1) well anterior to the upper jaw. No indication transformation and compared using analy­ of abilities to protrude the jaw was observed sis of variance (ANOVA) (GLM Procedure in any of the species, and no differences be­ [SAS Institute 1992]). Means were separated tween sexes were observed for this or any at P < 0.05 using Duncan's multiple range other trait. test. Longer gut lengths (larger ratios) are The five native stream gobioids differed typical of herbivores that ingest fibrous plant significantly in gut length relative to body foods that resist digestion, whereas carni­ length and in the pattern of intestinal con­ vores have shorter