16. Grazing Fishes as Components of North American Stream Ecosystems Effects of Campostoma anomalum 4- 2 A4 s William J. Matthews, Arthur J. Stewart, and Mary E. Power Abstract Algivorous fishes in North American streams have been largely overlooked by ecologists. Our work indicates that a widespread and abundant grazing minnow (Camposroma anomalum) can strongly influence distribution or standing crwp of attached algae in streams; they thus have the potential to significantly impact biological processes fundamental to stream ecosystems. Studies of this species in a southcentral Oklahoma prairie-margin stream (Brier Creek) showed that: (1) Camposroma can control the pool-to-pool distribution of algae, and (2) the presence of predators (Microprerus) influences the distribution of Campostoma. Other streams in the Ozark and Ouachita uplifts in Oklahoma and Arkansas showed very different patterns of bass-Campostom-algae distributions. In Ozark and Ouachita streams, Camposroma were larger, more abundant, and foraged more freely than in Brier Creek. Grazing by Campostoma in Ozark and Ouachita stream appears to have a strong influence on the kinds and standing crops of attached algae. We review the distribution of OW, potentially important algivorous North American stream fishes and outline certain interactions between grazing fishes Bpd# other ecosystem components that a priori are likely to be important. Recent reviews and conceptual syntheses of stream ecology (Vannote et al., 1980; Barnes and Minshall, 1983; Fontaine and Bartell, 1983) have illuminated the historical dichotomy between those stream ecologists who study fishes and those who do not. These important publications include only two chapters specifically addressing stream fishes, plus a few basic statements about longitudinal distribution patterns of fish in streams. Much contemporary literature in stream ecology ignores the fact that large, active, and abundant fishes may play important roles in system-level processes in streams. Most authors have failed to consider the role that herbivorous fishes can potentially play in influencing the distribution or standing crops of periphyton, the sites or rates of primary production, or the processing of organic matter, as observed for grazing or scraping invertebrates by Cum- mins (1978), Gregory (1983), and McAuliffe (1984). Important interactions occur between grazing fishes and their plant "prey" in marine habitats such as reefs (Lobel, 1980; Montgomery, 1980; Montgomery et al., 1980; Sam- marco, 1983; Meyer and Schultz, 1983, sea-grass beds (Weinstein et al., 1982), intertidal zones (Lubchenco, 1982; Miller, 1982), and tropical freshwaters (Lowe-McConnell, 1975; Goulding, 1980; Power, 1983, 1984b). Hynes (1970) described much of the basic ecology of stream fishes but included little information on the ecology of grazing fish Several studies have suggested the importance ol carnivorous or detritivorous fishes at the ecosystem level a North Temperate waters (Juday et al., 1932; Hall, ln Rickey et al., 1975; Durbin et al., 1979). Neverthela native grazing fishes have remained largely ignored as colp ponents of temperate stream ecosystems; most studies LI organic-matter processing continue to focus on activities a invertbrates, fungi, or bacteria. At least 42 species of stream fish in North America largely herbivorous; 24 of these species feed predominao? by grazing attached algae (Lee et al., 1980). Several sPerr, (e.g.. Cyprinodon spp. and Gila spp.) occur only in spring@ sinkhole systems in the western United States. Howevs. these restricted species are often abundant where they and can rival invertebrates as consumers of attached algaea higher plants. Some large, abundant catostomids that fl primarily herbivorous (e.g., Carostomus santannae, whd occurs in four river systems of southern California) c@ comprise a large part of the biomass of certain west@ I streams. Many species of algivorous minnows occur in the easw half of the United States. Some, such as NorropispilsbrYid d Notropis cerusinus, occasionally have large quantities I 128 Grazing Fishes as Components of North American Stream Ecosystems 129 damentous algae in their guts (Matthews et al., 1979; Surat al., 1982), although in such cases algae likely are con- umed incidentally with aquatic insects. Others, such as ;orropis nubilus. Phoxinus erythrogaster, and Phoxinus umberlandensis, feed almost exclusively on algae, which hey scrape from the surfaces of rocks. Several of these ninnows are among the most abundant fish species in their cspective habitats. The most abundant and widespread algivorous minnows in lastern and central North America are in the genus Campo- ioma. Campostoma anomalum ranges across more than half If North America (Burr, 1980a). Campostoma oligolepis rcurs in the uplands of the Midwest and in the Tennessee tiver and Mobile Bay drainages, and Cwnposrom paucir- ldii is found in eastern Gulf Coast drainages (Burr and Zashner, 1983). Campostoma anomalum is a large minnow 10 230 mm [SL]) that can be extremely abundant in small or nedium-size streams, sometimes dominating the fish com- nunity numerically and on the basis of biomass (cf. Lennon md Parker, 1960; Beets, 1979). In one pool (about 100 m long) in an Ozark stream, for example, our snorkeling sur- qs have provided estimates of more than 5,000 C. ano- wlum, many more than 120 mrn total length. In this chapter we summarize our studies of the ecology of C. anomalum and attached algae in streams of the Midwest. Ne recapitulate studies involving interactions between the mer and its food and describe a three-trophic-level interac- fnm in which predators (bass) control distributions of Cam- Psfoma and, indirectly, attached algae. We also present new data providing a broader view of predator-Campostoma- dgae relationships in several midwestern streams and offer aplanations for differences in the patterns observed in dif- knt stream systems. Finally, we speculate from pre- ha, data about the ecological significance of Camposto- (and other grazing fishes) in North Temperate stream -systems. bmpostoma Anomalum in Brier Creek, Oklahoma her Creek, in south-central Oklahoma, is a small prairie- Wpin stream of the Red River drainage. A 1-km mid-reach -[ion consisting of a series of 14 pools and their attendant contains a substantial number of Campostoma ano- Wh: this section of stream serves as the site for studies by * research group. Power and Matthews (1983) noted that Campostoma in- &nced standing crops of attached algae in this stream eh. Four of the 14 sequential pools in the study site Qbined large schools (50-400 individuals) of Camposto- ad lacked large predatory fishes. These "Campostoma his" were essentially devoid of algae except along shallow Xd mugins. Nine other pools lacked Campostoma and *Qined 3-8 large (> 70 mm SL) largemouth (Micropterus *Oides) or spotted (Micropterus punctufatus) bass. All "bass pools" had prominent growths of filamentous ?n algae (primarily Spirogyra sp. and Rhizoclonium sp.). % and Campostoma co-occurred in only one relatively qe Pool. Bass occupied deeper areas of the pool, and `*posroma occupied shallower areas. Within this pool attached algae occurred in deep areas where bass patrolled but not in shallow areas occupied by Campostoma. As far as we know, this was the first indication in a freshwater stream that piscivorous fish could influence distribution of a smaller herbivorous fish, which in turn influenced the distribution of attached algae. We (Power and Matthews, 1983) transferred various algae-covered substrates from bass pools to a Campostoma pool to assess short-term grazing impacts of Campostorna. Campostomu immediately swarmed to and actively fed on all these new algae-covered substrates but virtually ignored bare control cobbles relocated within the same pool. Algal biomass on the transferred substrates was dramatically and rapidly reduced; in 24 h ash-free dry weight of algae on the cobbles was < 25% of its initial value. In the initial study we also showed that bass could alter the use of habitat by Campostoma. Algae-covered substrates were again transferred into a Campostoma pool, and these substrates were actively grazed by these minnows. A large- mouth bass (300 mrn SL) was then tethered near the algae- covered rock with a line attached to its lower jaw. Over the next four days the tethered bass effectively "guarded" the transferred algae-covered cobbles; similar cobbles placed 1.3 m away were grazed frequently by Campostoma. Five other snorkeling surveys of fish in the Brier Creek study reach were made in 1983. The occurrence (presence- absence) of bass and Campostom in the 14 pools was in- versely related in all but one of these censuses (Power et al., 1985). Bass and Campostoma co-occurred in more than 2 of the 14 pools on only two occasions-after major floods. In late summer and in both autumns of the study algae and Campstoma were inversely distributed in the pools. Howev- er, during spring some algae (predominantly Spirogyra) accumulated despite grazing by Campostoma and became conspicious in most pools. Algae were also intermittently scoured from most stream substrates by large floods. During and immediately after periods of high discharge, then, algal standing crop may have little relation to distribution of grazers (cf. Fisher et al., 1982; Power and Stewart, in press). However, during extended periods without floods (i.e., "normal" low-flow regimes). grazing by Campostoma in Brier Creek establishes a clear and recurrent pattern: there is a markedly lower standing crop of algae in Brier Creek "Campostoma" pools than in pools that lack Campostoma. Grazing by Campostoma can apparently regulate standing crops of attached algae, even when Campostoma densities are somewhat less than normal. In autumn 1983 a pool containing bass and large quantities of attached algae was split longitudinally by a plastic fence, and we removed the bass and all large sunfish by electroshocking (Power et al., 1985). We stocked Campostoma on one side of the pool (at a density slightly less than that found in most Campostoma pools) and left the other side of the pool (and another, nonmanipulated Cumpostoma pool) as a control. Over the next five weeks attached algae on the Campostom side of the pool declined rapidly and remained low for the duration of the experiment. On the side of the pool lacking Cumposro- ma, a bloom of algae (largely Spirogyra) occurred. During the experiment filamentous green algae were conspicuous in 130 Community and Evolutionary Ecology of North American Stream Fishes other pools of the reach that lacked Campostoma, but not in pools containing Campostoma. Two other experiments in 1983 and in 1984 evaluated the impacts of predators (largemouth bass) on the distribution and use of habitat by Campostoma. In these experiments (Power et al., 1985) we added largemouth bass to a pool containing a school of Campostoma. In both experiments addition of bass resulted in immediate changes in habitat used by Cumpostoma: some emigrated from the pool, and others moved to shallow-water areas near the margin of the pool. Campostoma remaining in the pool after bass were added spent significantly less time feeding and more time hiding among cobbles than they did in the pool before bass were added. Substantial regrowth of attached algae occurred one to two weeks after addition of bass in both experiments. Collectively, our observations and experiments in Brier Creek indicate that during periods of normal flow (1) Cam- postoma can regulate standing crops and distribution of attached algae and (2) bass can influence distribution of Campostoma, both within and among pools. The results from the Brier Creek study suggested two broader questions: First, are the patterns evident in Brier Creek typical of those in other stream systems in the Midwest? Second, does graz- ing by Campostoma have predictable effects on stream algae and stream ecosystem processes? We address these questions below. Stream Surveys of Bass-Campstoma-Algae In September 1983 and in June-July 1984 we quantified distribution of bass, Campostoma, and algae in two Ozark uplift streams and in a stream in the Ouachita Mountains of LeFlore County, Oklahoma (table 16.1). In April 1984 we conducted a similar survey in Brushy Creek (Johnston Coun- ty, Oklahoma), and in July 1984, we surveyed Tyner Creek (Adair County, Oklahoma). In all of the surveys Matthews snorkeled slowly upstream through each pool, recording numbers of Campostoma and numbers and species of Mi- cropterus; sizes of all bass were estimated to the nearest inch total length (TL). Algal height and composition were de- termined by Stewart and Power (September 1983-April 1984), who used methods described in Power and Matthews (1983). In July 1984 algae were asssessed more rapidly by scan surveys. In each pool numbers of invertebrates were estimated by having one person spend 10 minutes picking organisms from all available kinds of substrates with forceps. In Brushy and Tyner Creeks (table 16.2) so few bass were present that these streams appear to represent Campostoma distribution in essentially predator-free environments. In these two streams Campostoma were present in relatively low numbers but were widely distributed in almost all pools. One exception was that no Campstoma occurred in the four most densely shaded pools of Tyner Creek where substrates on the stream bed were virtually devoid of algae. In all other surveyed pools at least thin, slick coatings of epilithic di- atoms occurred on submersed rocks. In the four canopied pools of Tyner Creek rocks were not even slippery to the touch, suggesting that Campostoma may avoid areas of ex- tremely low primary productivity. One Ozark stream (War Eagle Creek) and the Oua& Mountain stream (Big Eagle Creek; table 16.2). had g termediate densities of bass (Micropterus > 150 mm TL 17-50% of all pools; some pools had as many as 8 bass). 8 these two streams, however, Campostoma were present virtually all pools, including those with bass (table 16% The second Ozark uplift stream (Baron Fork, in nom Oklahoma) had the largest number of bass per pool that observed in these stream surveys (table 16.1). Despite & presence of 20 or more bass in some Baron Fork pooh, however, Campostoma remained abundant. Even a inspection of table 16.1 shows that the strong "Brier Cree& type" of pool-to-pool complementarity of bass and Camp stoma does not occur in streams we surveyed in the Om& and Ouachita mountains; there bass and Campostoma distri. butions among pools were independent (x2 = 0.942; p 0.33; 2 x 2 contingency analysis). Of 86 pools surveyd outside Brier Creek, Campostoma occurred without bass i 51 pools but co-occurred with bass (> 150 mm TL) in 29 At least three factors could account for the lack of 1 bass-Campostoma complementarity pattern in streams of tbc Ozark and Ouachita uplands: (1) pool size, (2) ease of move ment across riffles between pools, and (3) aspects involviq prey size and/or predator capabilities. Some Ozark-Ouachta stream pools may simply be too large to be effectively p, trolled even by several bass. In larger Ozark pools bass oftea occupied deeper portions of the pool, while schools of Cm postoma remained near the substrate on gravel slopes rt distances > lm from larger bass. Riffles connecting Ozark pools are deep (often 1615 cm) compared to those in Brier Creek (2-4 cm). Hence mo4 riffles in Brier Creek can preclude movement of large like bass at normal flow; but Campostoma and bass readily traverse most riffles in the Ozark-Ouachita streams. Formation and maintenance of bass-Campostoma corn plementarity in Brier Creek may therefore depend in part m restricted movement of fish between pools (which allow time for attrition of Campostoma) or on more frequent pool- to-pool movement of prey (Campostoma) than of predators (bass). The specific type of predator and size of prey may also influence bass-Campostoma interactions in Ozark stream- Most bass in Ozark and Ouachita streams are smallmouth bass (Micropterus dolomieui) (table 16. l), which have re- latively smaller mouths than largemouth bass. AdditiondY- smallmouth bass typically eat crayfish more often than fish (Lewis and Helms, 1964; Carlander, 1977), althoughtheydo feed on a variety of minnows (Scott and Crossman, 1973)- this context it may also be important that Campostom virtually all the Ozark and Ouachita streams are larger tha those in Brier Creek. To date we have used nondesUuctiVe census methods and thus lack precise length-frequency da' for Camposroma in these streams. However, the same In- dividual (Matthews) has made all snorkeling observations- standardizing the estimates to a large degree. Campostorno larger than 150 mm TL are very rare in Brier Creek; largest individuals we have collected there are about 125 fl TL. In Ozark streams schools of Campostoma (often 100- 500 individuals) frequently consist of individuals l25-15O pools. ! ab / ,.^. .. C( i - -J i I .. ,\ '.\ OB Is0 lth re- ly, sh do In in 111 ve .la n- S, la Xi. le a )- 0 Grazing Fishes as Components of North American Stream Ecosystems 131 Table 16.1. Numbers of Campostom (Minnows). Bass (by Size, Class, TL), and Invertebrates in Pools of Streams in Oklahomaand Arkansas strcam (Date) Pool Taxa 1 2 3 4 56 7 8 9 10 11 12 13 14 - Brushy Creek (April 1984) Campostom SMB* 150-250 mm SMB%250 mm LMB'2150 mm Invertebrates Tyner Creek (July 1984) Campostom SMB 150-250 mm SMB 2250 mm LMB 3150 mm Invertebrates War Eagle Creek Campostoma SMB 150-250 mm SMB 2250 mm LMB 2150 mm Invertebrates (September 1983) War Eagle Creek (June 1984) Campostoma SMB* 150-250 mm SMB a250 mm LMBt 2150 mm Invertebrates Big Eagle Creek Campostoma SMB 15&250 mm SMB 3250 mm LMB 2150 mm Invertebrates Bicn Eagle Creek (July 1984) Campostoma SMB 150-250 mm SMB a250 mm LMB 3 150 rnrn Invertebrates (September 1984) 0 1 18 540 25 630 t 210 9 112 3 2 51 186 3 9 9 19 32 31 110 13 1 16 21 0 800 t t 33 22 10 8 70 50 48 30 60 36 13 58 1 39 104 11 503 t 202 3 4 14 50 45 50 12 17 10 19 20 * 24 23 17 0 25 13 32 3.000 500 6 t t 210 1,820 120 2 11 6 13 60 550 6 1 2 41 51 277 370 105 7 2 1 37 27 21 30 22 0 33 765 17 88 31 32 20 0 t 310 9 44 29 82 39 0 t 25 11 250 35 5 21 23 1 * 109 12 58 2 25 1 53 655 2 1 6 19 % 9 t 254 19 37 23 1 30 Sore. Numbers of invewbrates = numbers of individuals collected in 10 minutes of picking with forceps by OM investigator. YSMB = smallmouth bass. . = largemouth bass. 'VQ invertebrate sample taken. i t 132 Community and Evolutionary Ecology of North American Stream Fishes -4 -# Table 16.1. Continued + Stream (Date) PO01 Q Taxa 1 23 4 5 6 7 8 9 10 11 12 13 14 ~ Baron Fork (September 1983) Campostoma 800 1,000 500 232 2,450 800 50 5,350 3,000 SMB* P 150-250 mm 73 1 38 6 20 17 SMB 3250 mm 6 7 LMB+ 2150 RUII 2 Invertebrates 76 95 107 111 136 134 70 99 91 Baron Fork (July 1984) Campostoma 1,003 1,ooO 350 1 1,920 1,020 550 16 1,195 120 76 1,490 385 1,210 SMB 150-25Omm + 13 6 5 2 14 5 1 SMB a250 mm 8 1 LMB 2 150 mm Invertebrates 7 19 18 44 19 46 30 32 42 20 33 25 30 3 mm TL, and specimens as large as 160-175 mm are not uncommon. Smallmouth bass, therefore, may be less effec- tive predators on the large Cumpostomu in Ozark streams than are largemouth bass that prey on Curnpostoma in Brier Creek. In feeding experiments (April 1985) medium-sized smallmouth bass (ca. 250 mm TL readily ate Cumpostornu as large as those typical for Brier Creek. Cempostome Behavior Regardless of the ultimate causes, Cumpostomu in Ozark streams occur in most pools, where large individuals move and feed with little apparent restriction by predators. Un- derstanding the feeding ecology of Campostornu permits greater insight into the potential consequences of such un- restricted grazing. Campostomu typically feed in large schools, even at temperatures as low as 7" C (table 16.2). In Brier Creek pools with slow flow of water, Cumpostornu school when un- disturbed. They browse substrates, but schools move little. In larger pools and in shallow areas of Baron Fork pools where flow is greater, schools of Campostornu exhibit a distinct pattern of grazing and movement. A school of about 200-500 individuals, for example, often grazes on algae attached to cobble or gravel. Individuals are typically about 10 cm apart, and most orient upstream. The school as a discrete unit often grazes in a given area for 1-2.5 min; members of the school then cease feeding in unison, drift 3-5 m downstream, and resume grazing. A series of such grazing-drifting sequences continues until the school is dis- placed a considerable distance downstream. Eventually fish in the school move upstream, and the entire grazing-drifting sequence is initiated again. The pattern is not always as consistent as described; some- times Cumpostornu schools browse upstream into the current or show little net movement. However, repeated observa- tions suggest that (1) the schools have discrete grazing pat- terns, (2) schools graze relatively large areas each day, and (3) within-pool grazing movements may depend on pool size, depth, and water veIocity. The extent to which Camp stoma grazing and movement depend on availability of algae or on predators, or what cues influence their behavior, re mains unknown. Cumpostomu have distinct feeding modes that appear to bc influenced by availability of particular types of food. 'Ihc most common feeding mode in Brier Creek, where Camp stoma often forage in deposits of detritus or on epiphytes growing on Churu, is "nipping"; this mode is similar to that described for Poeciliu (Dussault and Lamer, 1978). In nip ping, Cumpostornu incline their bodies at about 45" to the substrate and take small, rapid bites. At times several in- dividuals may take turns nipping at one small location on the substrate, suggesting that they are selective and that they receive foraging cues from each other. Detritus in Brier Creek "Cumposromu pools" contains few high-quality food items (intact algal cells), and these pools also have a low algal standing crop. In such conditions selectivity would be beneficial. "Swiping" is the most common feeding mode displayed by Cumpostornu in Ozark streams. This mode is used 10 detach blue-green and diatomaceous algae growing as dense, 1- to 2-mm-thick felts on rock surfaces. In swiping, Campo- stoma position themselves above the substrate, hold their bodies rigid, manuever with rapid pectoral fin beats, and then strike the algae suddenly with sharp sideways thrusts of the head so that the cartilaginous lower jaw scrapes algae from Table 16.2. Feeding Rates of Campostoma* Date Temp. ("C) Location SL(mm) X SE N' Nov. Nov. Dec . Nov. June ~~~~~~ 1982 17 Brier Creek 50-60 16.7 1.5 28 1982 17 Brier Creek 5&60 8.9 1.9 21 1982 7 Brier Creek 50-60 11.2 2.1 10 1982 10 Baron Fork 50-130 15.2 3.1 17 1984 25 Brier Creek a80 10.8 2.3 35 *Bites per minute `Number of individual fish observed and timed. # A Grazing Fishes as Components of North American Stream Ecosystems 133 I ! rocks. This feeding mode is physically more vigorous . nipping. Swiping leaves distinct grazing scars on rock, ; .d, or leaf substrates (Matthews et al., 1986). Often a fish des two "swipes" in rapid succession, leaving a pair of ping scars. Grazing scars are common where schools of k cmpostoma have fed. t .A third grazing mode, "shoveling," is used by Camposto- ,,,,I when they feed on algae attached to large, smooth rock d-aces. In shoveling, Campostom push their lower jaws against the substrate and swim forward, removing algae en mte. Campostom feeding in this manner make grazing trails several cm long; these trails are readily distinguished from the more meandering trails left by snails. Kraatz (1923) observed similar behavior of Campostoma feeding on di- 310maceous mats on aquarium walls. \\;hen grazing on blue-green algal felts, Campostoma do not remove algae down to the bare rock surface. Microscopic examination of typical scars indicates that the fish remove only the upper layers of algae and on epialgal layer of mucilaginous material heavily invested with bacteria. A thin film of algae (about 10% of the pregrazed crop) typically remains within the scars. The grazing scars left by Campostoma often cover much of the available rock substrates where schools have recently worked. We (Matthews et al., 1986) quantified the number and sizes of grazing scars left by Campostoma in a pool with bedrock substrate and predominantly blue-green algal felts (Tyner Creek, Oklahoma). At depths of 10 to 59 cm the number of grazing scars attributable to Campostoma aver- aged 1,800 per m2; these data included numerous sites in water < 20 cm deep, where there were few or no scars. At depths > 40 cm we found > 5,000 grazing scars perm'. The size of individual scars was highly variable but averaged 0.57 em2 (N = 90). Thus across all depths in this part of Tyner Creek about 10% of the submersed substrate area exhibited evidence of grazin by Camposroma. In deeper areas with > 5.000 scars per m5, as much as 28.5% of surface areas of substrate was, on average, recently grazed. Consequences of Herbivory by Campostoma Huge schools of Campostoma, some containing thousands of individuals, are common in some Ozark streams. Such Streams show evidence of intense grazing by Campostoma; the grazing scars produced by these minnows often coalesce, covering most of the surface area of heavily grazed sub- strates. In Baron Fork of the Illinois River, Campostoma are extremely abundant, and the standing crop of algae is con- sistently low. Epilithic communities of algae in this stream are primarily slick, dark-colored blue-green felts 1-2 mm thick. Although thin, these felts are highly productive (0.6 g Odm'/h; Stewart, unpub. data), suggesting that the relatively low standing crop of algae (108-325 g dry weight/m2) is more likely the result of biomass removal by grazing than of low rates of algal growth (cf. Gregory, 1983). On the basis of Our censuses and observations to date, Campostoma appear to be the major herbivores in this and many other Ozark Mountain streams. What are the ecological consequences of intense herbivory by stream fishes? Consequences of Herbivory by Campostoma to algae In October 1984 we incubated glazed ceramic tiles in Fiber- glas troughs in Baron Fork. The troughs permitted substan- tial through-flow of water but excluded Campostoma, snails, and crayfish. Chironomids gained access to the tiles, but their characteristic grazing pattern (a circular area cleared around their point of attachment) permitted areas they affected to be readily identified. In the stream outside the troughs, where Campostoma grazed heavily, the typical algal flora consisted of a 1- to 2-mm-thick felt of epilithic blue-green algae (largely Calothrir, Phormidium and Oscil- latoria spp.). Tiles incubated within the troughs rapidly developed a diatomaceous flora dominated by Melosira, Cymbella and Synedra spp. (Power et al., unpub. data). When these tiles were moved to the stream bed, they were actively grazed by Campostoma; the diatoms turfs were within weeks replaced by blue-green felts. Conversely, natu- ral felt-covered slate substrates transferred from the stream bed into the troughs were overgrown by diatoms within 410 days. Campostoma finally gained access to the ends of the troughs (because of sagging of the ends) and grazed on algae growing on the tiles. These grazed tiles also developed blue- green felts, thereby ruling out "trough effects" as a cause of differences noted for substrates exposed to and protected from grazing by Campostoma. These results show that (1) grazing by Campostoma can maintain low standing crops of algae on rocks of Ozark streams, and (2) grazing by Campo- stoma alters taxonomic composition of the algal community that develops. That herbivory by Campostoma has the potential to regu- late standing crops or kinds of stream algae during much of the year is suggested by all of our studies in Brier Creek and in streams of the Ozark region. Because grazing by Campo- stoma typically removes mainly the algal overstory, light and nutrients are more available to algae remaining within the grazing scars. Moderate intensities of grazing by Camposto- ma therefore stimulate primary productivity per unit algal biomass. Additionally, microscopic examination of Cam- postoma feces shows that some fraction of the ingested algae pass succcessfully through the gut and remain viable. Although we have some data suggesting that epiphytic di- atoms suffer greater mortality than their filamentous green algal hosts in passage through the alimentary canal of Cam- postoma, additional data are needed before we can determine the consequence to algae of passing through the gut of Cam- postoma. Campostoma could conceivably benefit some filamentous algae (e.g., Rhizoclonium) by "stripping" them of encrusting epiphytes (cf. Lubchenco, 1983). Campostoma feces contain both "regenerated" nutrients and viable algae; the feces accumulate in large quantities in deeper areas of all Brier Creek "Camposroma pools," and are reworked frequently by Campostoma when algae are scarce. Nutrient translocation as a result of fish activity is important in reef systems (c.f. Meyer and Helfman, 1983; Meyer and Schultz, 1985) and has been shown to be impor- tant in freshwater streams (Hall, 1972; Durbin et al., 1979). The significance of the accumulation of Camposroma feces to the ecology of algae that are not consumed or to those that are consumed but survive gut passage remains unknown. A 134 Community and Evolutionary Ecology of North American Stream Fishes To date we know little about relationships between Cam- postoma and the spatial distribution of algal biomass and productivity. In a tropical stream Power (1983, 1984) showed that densities of grazing catfish tracked algal pro- ductivity; catfish were seven times more dense in sunlit pools, where primary productivity was 6-7 times higher than in dark pools. Our data from Tyner Creek suggest that Cam- postoma are absent where extensive canopy cover limits algal growth. Although removal of algae by grazing and subsequent recovery of algae in previously grazed areas undoubtedly affects the way stream habitat is used by Cam- postoma and other stream algivores, data addressing these aspects are still unavailable; much remains to be learned about dynamics of grazing fishes and the distributions of algae in North American streams. Consequences to Other Stream Biota Periphyton communities provide refugia for a variety of invertebrates (Cuker, 1983), and removal of erect or foliose algae has important consequences to stream invertebrates. We found relatively similar numbers of invertebrates in pools with few or many Campostoma in the Ozark and Ouachita streams we surveyed (table 16.1), but the scale of our observations or our collecting techniques may have been inadequate to detect differences. Herbivory by Campostoma might influence stream invertebrates in two ways: first, graz- ing by Campostoma alters the standing crop and growth forms of algae and so alters the types or amounts of food or shelter available to grazing invertebrates; second, Camposto- ma could promote downstream drift of stream invertebrates if their disturbance of algae and rock substrates causes in- vertebrates to enter the water column. If Campostoma re- move algae that harbors invertebrates or increases their emigration, a given stream reach may support fewer in- vertebrates and, therefore, be a less profitable place for insectivorous minnows to forage. Our snorkeling observa- tions suggest that stream reaches containing many Campo- stoma have a lower diversity of other small fishes (compared to typical stream segments not dominated by Campostoma), but additional data in a range of stream types are needed to evaluate quantitatively the impact of high densities of Cam- postoma on fish-community structure. Positive interactions between Campostoma and other fishes could also exist. For example, our snorkeling observa- tions in numerous Ozark streams reveal that Campostoma and Notropis pilsbryi often occur in the same pools, in very close proximity or actually intermixed. Feeding opportuni- ties for N. pilsbryi may be improved by the grazing activities of Campostoma, the former consuming invertebrates made available in the water column by grazing activities of the latter. Consequences to the Stream Ecosystem Grazing by Campostoma initiates a cascade of effects appar- ent at both large and small spatial scales. At the whole-stream level these effects can be expected to include large-scale changes in patterns of nutrient uptake, regeneration, and downstream transport; changes in overall rates of primary production and in the sites and rates of decomposition; 4 increases (or decreases, depending on flow regimes) in & degree of spatial heterogeneity of various biotic procew We offer here speculations about the possible range and tyPg of processes and conditions that herbivory by Campost- may initiate and mantain in stream ecosystems. streams such as Baron Fork schools of Cmposroma gener& and maintain "grazing lawns" (sensu McNaughton, 1984), much as herds of ungulates do on the Serengeti grass14 The algal lawns maintained by Campostoma in O& streams consist largely of tightly attached blue-green algr' (notably Calothrix) that have prostrate growth forms; lk grasses, Calothrix has a "basal meristem" (most cell divj. sions occur just above the basal heterocyst; B. Whitton, University of Durham, England; pers. comm. with Stewart). This spatial "growth refuge" allows Calothrix to persist despite intense grazing by Campostoma. The Calothrix lawn that forms in response to grazing by Campostoma can bc expected to alter large-scale nitrogen cycling characteristics of the stream, for Calothrix fixes N2, while diatoms (which flourish when Campostoma are excluded) do not. Because many blue-green algae can fix Nz and because nitrogen content of food is sometimes an important determinant of food quality for herbivorous invertebrates (Ward and Cum- mins, 1979) and fish (Horn et al., 1982; Horn and Neighbors, 1984), Campostoma may indirectly influence growth rates or use of space by other stream algivores such as crayfish, snails, and aquatic insects (see also Hart, 1985; McAuliffe, 1984). Changes in algal community composition owing to graz- ing by Campostoma may also have consequences to tbe cycling of nitrogen and carbon at very small spatial scales. Microscopic examination of the blue-green algal commuN- ties dominating when Campostoma were present showed that many bacteria were attached to the mucilaginous sheaths of Calothrix; diatoms dominating when Campostoma were ex- cluded supported visibly fewer bacteria. Jones and Stewart (1969) found that Calothrix scopulorum released combined nitgrogen compounds that were readily assimilated by vari- ous fungi and bacteria; other filamentous blue-greens do so also (Paerl, 1978). We do not yet know the extent to which bacteria influence nutrient and energy fluxes in Baron Fork, but Cole (1982), Newbold et al., (1983), and Cume and Kalff (1984) show that this possibility should not be over- looked. The effects of Campostoma on nutrient cycling may change seasonally. In early spring, for example, Campost@ ma often grazed on bedrock substrates in shallow, fast- flowing areas of Brier Creek; these substrates supported visually uniform thin layers of attached diatoms dominated by species of Synedra, Gomphonema, and Cymbella. Cam- postoma feces were displaced from three areas by water currents and accumulated downstream in crevices and deeper pockets in microdepositional zones. Water level and velocity in Brier Creek decline in summer, and most riffles become SO shallow that Campostoma no longer feed there. Campostom then graze almost exclusively in deeper areas of pools; and owing to lack of downstream transport, their feces accumu- late where feeding occurs. Over larger spatial scales herbi- 4 Our observations and experiments imply that in 5 of ex- /art led ui- ich rk, .nd er- aY `0- st- ed :d n- er 2r !Y so ;0 `U d I- t- Grazing Fishes us Components of North American Srream Ecosystems 135 ,on by these minnows in summer lowers spatial heterogene- .n if algae within pools ("Ccunpostoma pools" are uni- low in algae) but increases heterogeneity of algal biomass (and possibly growth) between pools ("Cumposro- pools" versus "bass pools"). By late summer most **Cornpostoma pools" are so nearly devoid of algae that bare rubstrates predominate in sites that are frequently grazed; bye sites will alter sediment-water exchange characteristics for most nutrients (cf. Mulholland et al., 1983). When primary productivity is nutrient-limited and the rate of regeneration of nutrients from Cumposroma feces is low, ustamed grazing may gradually lower productivity and dis- ion normal patterns of nutrient spiraling in streams by con- wting nutrients into inaccessible forms. Conversely, if dgae are nutrient-limited and regeneration of nutrients from Camposromu feces proceeds rapidly, moderate grazing may increase productivity by favoring algae with higher rates of turnover (see also Mulholland et al., 1983; and Horn, 1982). In streams where algae are not nutrient-limited, grazing by Cumpostoma could increase primary productivity by remov- ing algal "overstories" that reduce light to understory com- munities or by altering feeding behavior of other algivores. Conversely, they may decrease total productivity on an areal basis if they remove excessive quantities of algal biomass or if they promote the development of algal lawns comprised of species with lower productivity but which persist by virtue of having polar cell division (such as Gloeorrichiu or Calothrix). In general, intense grazing by Cumpostomu will alter spatial distributions and relative intensities of biotic (algae and bacteria) and abiotic (exposed sediment) pro- cesses controlling nutrient uptake and release. The con- sequences of such changes to nutrient-spiraling characteris- tics at very large spatial scales (km reaches) remain un- known. In the arguments above, a very fundamental question remains unanswered: How much of the ecological theory generated from studies of grazers in low-vectored terrestrial or lake systems (cf. Noy-Meir, 1975; McNaughton, 1979, 1984; McNaughton et al., 1982; Caughley and Lawton, 1981) can be applied directly to grazing processes in streams that include some factors that are strongly vectored by the flow of water? In aquatic ecosystems many of the obvious ecological consequences of grazing are immediately nutri- ent-related, because nutrient availability limits productivity in aquatic ecosystems more often than in terrestrial ecosys- tems (cf. Ricklefs, 1979). Nutrients released by stream herb- ivores can be effectively carried away from sites of nutrient uptake by the flow of water. In summary, our studies to date suggest that Cumposromu, as abundant, active algivores, may strongly influence a vari- ety of very basic processes in North American streams. As yet we have investigated only a few aspects of the ecology of Cumposromu in detail; other features about its ecology for which we have only preliminary data further suggest its potential influence in biotic communities and in stream ecosystems. Our work has focused on a single species of large herbivorous minnow, but many other abundant or widespread fish taxa may play similarly important roles in streams throughout the United States. We hope that ichthyologists and stream ecologists in general will be stimu- lated to consider in more detail the potential significance of grazing fishes in various stream ecosystems.' ' We thank Bret Harvey, Beth Goldowitz, Larry Greenberg, Sheila Wiseman, Stephanie Contreras, Arlene Stapleton, Scott Matthews, and Tom Hegerfor help in the field or laboratory and Bob Cashner, Bill Dietrich, and Steve Threlkeld for helpful discussions at various phases in our work. We are particularly grateful tothe Panish family, Marshall County, Oklaho- ma, for access to field sites, and to Phil1 and Donna Bright and the staff of Camp Egan. Oklahoma, for many courtesies. Susan Kelso typed innumer- able drafts of the manuscript. This research was supported initially by a visiting investigator award to Matthews and Power from the Department of Zoology, University of Oklahoma, and has been funded since October 1983 by the National Science Foundation (BSR-8307014).