Nonh American Journal of Fisheries Management 18:274-285, 1998It> Copyright by the American Fisheries Society 1998

Ecology and Status of Piscivores in Gun, an OligotrophicTropical Reservoir

JOHN D. WILLIAMS' AND KIRK O. WINEMILLER*

Department of Wildlife and Fisheries Sciences, Texas A&M UniversityCollege Station, Texas 77843, USA

DONALD C. T APHORN

Universidad Experimental de Los Llanos Ezequiel ZamoraGuanare, Estado Portuguesa 3310, Venezuela

LUIS BALBAS

Corporaci6n Venezolana de Guiana-Electrijicaci6n del Caron(Departamento de Estudios Basicos

Apartado Postal 28. Puerto Ordaz. Estado Bolivar 8015. Venezuela

Abstract.-Guri Reservoir in Venezuela supports sport fisheries for two cichlid species (pavonvenado or speckled pavon Cichla temensis and pavon mariposa C. orinocensis) and payara Hy-drolycus scomberoides. The lake receives acidic, hyperoligotrophic "black waters" from the CaroniRiver and experiences water level fluctuations associated with seasonal precipitation and hydro-electric operations. During the early 1980s, piscivore populations expanded in the new reservoir,but the quality of sportfishing declined during the 1990s. During 1993-94, we studied the ecologyof the two Cichla species, H. scomberoides, and a fourth piscivore, Plagioscion squamosissimus(curvinata or silver croaker) and compared our data with those from similar surveys performed7-9 years earlier (0-2 years after attainment of the reservoir's current crest height). Seine andgill-net samples from the lake's northern region produced 50 fish species from 18 families. Seinesamples were dominated by a small characid, Hemigrammus micropterus, whereas gill-net sampleswere dominated by a large detritivorous characiform, Prochilodus rubrotaeniatus. Gill-net catchrates were low in all three surveys. All four piscivores from our recent survey had better bodycondition than in a 1985-1986 study, an indication of possible growth compensation. Diet breadthsof all piscivores were low, and diet overlap was low for H. scomberoides with the two Cichlaspecies and for P. squamosissimus with C. orinocensis. Hydrolycus scomberoides and C. temensisconsumed significantly larger prey than P. squamosissimus, and differences in the relative pro-portions of specific characid and cichlid fishes consumed by the two Cichla species indicate thatthey forage at different depths within the littoral zone. The combined effects of hyperoligotrophicwaters entering the reservoir, reduced inputs of dissolved nutrients from submerged terrestrialvegetation, low retention time of the reservoir, and harvest probably influence the low abundanceof the two Cichla species in Guri Reservoir.

Most information on reservoir limnology andfisheries comes from temperate regions, where im-poundments, including those created for hydro-electric generation and flood control, have en-hanced recreational fishing opportunities (Prosser1986). In recent years, construction of large res-ervoirs has slowed in the United States, due, inpart, to increased awareness of environmental andeconomic costs. However, hydroelectric facilitiescontinue to be constructed on a grand scale in otherregions, especially the neotropics (Fearnside 1989;Allan and Flecker 1993). Large hydropower im-

poundments can directly affect tropical fisheriesthrough alteration of hydrology, habitat, and mi-gration routes (Barthem et al. 1991). Some fluvialspecies disappear from reservoirs, whereas otherspersist at altered densities (Alvarez et al. 1986;Winston et al. 1991; Agostinho and Zalewski1994). Reservoirs typically support fewer fish spe-cies than their associated rivers, often as a resultof large-scale changes in the regimes of temper-ature, turbidity, flow, allochthonous nutrient in-puts, and availability of food resources (Alvarezet al. 1986).

High productivity in new reservoirs has beenattributed to retention of allochthonous and au-tochthonous materials during initial impoundment,release of nutrients from inundated soils, andavailability of organic material from inundated

* Corresponding author: kowl956@zeus.tarnu.edu1 Present address: Prince William Sound Science Center,

Post Office Box 105, Cordova, Alaska 99514, USA.

274

PISCIVORES IN OURI RESERVOIR, VENEZUELA 275

strictions and catch limits are being developed fora limited-entry commercial fishery. These nego-tiations are being undertaken with scant infor-mation on the current p9pulation status of the twoCichla species and other large fishes. Despite thepopularity of Cichla spp. as sport fish and the sportfish potential of other predatory fishes, such aspayara Hydrolycus scomberoides, curvinata (or sil-ver croaker) Plagioscion squamosissimus, and red-eye piranha Serrasalmus rhombeus, little researchhas been conducted on fish ecology in Guri Res-ervoir. Initial assessments of the recreational andcommercial fisheries potential of the reservoir doc-umented 68 species, the most important (biomass)being Prochilodus rubrotaeniatus, Plagioscionsquamosissimus, and Hydrolycus scomberoides(Alvarez et al. 1986; Lasso et al. 1989; Novoa etal. 1989).

The four most common, large piscivores of GuriReservoir are Cichla temensis, C. orinocensis, P.squamosissimus, and H. scomberoides. The twoCichla species are nonindigenous to the Caron}River basin and were introduced into Guri whenponds were flooded during the final stage of res-ervoir filling during 1979 (Alvarez et al. 1986;Balbas and Pacheco 1989). Within 5 years, thesecichlids colonized the entire shoreline of the lake(Alvarez et al. 1986). Both Cichla species are na-tive in other rivers of the Orinoco River basin, andP. squamosissimus and H. scomberoides are nativeto the Caron} River Basin.

The goal of our study was to investigate anglerclaims of declining Cichla populations in Guri bysurveying the fish community in the northern re-gion of the reservoir. Our objectives were to (1)measure water quality; (2) assess the relativeabundances of littoral zone fishes that form theprey base for piscivores; (3) compare gill-net catchper unit effort (CPUE) of dominant species be-tween surveys conducted during 1985-1986, 1987,and 1993-1994; ( 4) evaluate population size struc-ture and length-weight relationships of the fourdominant piscivores; and (5) analyze diet com-position and interspecific diet overlap of the fourdominant piscivores.

substrates and vegetation (Prosser 1986). Follow-ing initial flooding, submerged organic matter de-composes and releases dissolved nutrients, whichusually increases production at all levels, includ-ing that of predatory fishes (O'Brien 1990). De-pending on local climate and the physical andchemical attributes of the reservoir, high rates ofproduction can last for many years during a periodknown as the "trophic upsurge" (Kimmel andGroeger 1986). As dissolved nutrients are eitherdepleted by the fauna (Goddard and Redmond1986) or flushed through the spillway, this highinitial fish production gradually decreases and lev-els off at a rate somewhere between those of nat-ural rivers and natural lakes of the same region(Noble 1986; Randall et al. 1995). This "boom-bust" cycle of reservoir production, which maylast 25-30 years, has long been recognized by res-ervoir managers in temperate regions (Rzoska1966).

Tropical reservoirs differ from temperate lakesbecause rates of physicochemical and biologicalprocesses tend to be faster at higher mean annualtemperatures. For example, the trophic upsurgemay last only 6-10 years in the tropics (Lowe-McConnell 1973). Water level fluctuations asso-ciated with hydroelectric facilities affect tropicalfishes, many of which spawn when water levelsrise instead of responding to variation in watertemperature or day length (Schwassman 1978). In-creasing demands for electricity in developingcountries have largely been met with hydroelectricpower, yet the full effects of river impoundmenton tropical river ecosystems are poorly under-stood.

Guri Reservoir (formed by Raul Leoni Dam),the eighth largest reservoir (4,250 km2) and thesecond largest hydroelectric facility in the world,currently supplies 70% of Venezuela's electricalpower (10,000 MW) (Morales and Gorzula 1986).During the 1980s, a sportfishing industry was es-tablished at Guri Reservoir, complete with hotels,boats, local guides, and tourist excursions. Theprimary sport fish species are two cichlid species(pav6n venado or speckled pavon Cichla temensisand pav6n mariposa C. orinocensis) that are locallycalled peacock bass or pavones. Although size (30cm minimum) and bag (daily possession of five)limits were established by the Ministry of Envi-ronment and Renewable Natural Resources(MARN), compliance and enforcement are poor atthe reservoir (L. Balbas, personal observation). Inrecent years, sport fishers have reported poorcatches. In addition to the sport fishery, gear re-

Study Area

Guri Reservoir is fed by the Caron} and Paraguarivers (Figure 1), both of which drain the ancientgranite outcroppings and nutrient-poor sandy soilsof the Guyana Shield region. Construction of GuriReservoir began in 1963, and a final crest heightof 272 m above sea level was achieved in 1986.Currently, the reservoir has a surface area of 4,250

276 WILLIAMS ET AL

63°00'W 62°30'W

..,.~~!)...\_--,--y~

VENEZUELA

FIGURE I.-Map of Gun Reservoir in Bolivar State, Venezuela, showing locations of sampling sites.

km2 and a volume of 135,000 X 106 m3 (Corpor-aci6n Venezolana de Guiana-Electrificaci6n delCaron! [CVG-EDELCA], unpublished report).

Annual rainfall at Guri is 256 cm (the majorityfalling between May and October) and annualmean water temperature is 26.5°C (Gonzalez et al.1989). "Black waters" (Sioli 1975) from the trib-utary rivers are characterized by low pH, low nu-trient content, low suspended particulates, andhigh concentrations of dissolved organic matterthat produce a dark, tea-stained appearance. Gurihas a relatively low retention time of hyperoli-gotrophic waters, so aquatic primary and second-ary production is extremely low (Gonzalez et al.1989; Weibezahn 1994). No true aquatic macro-phytes inhabit littoral areas in the lake's northernsection; however, aquatic plants were documentedduring the initial trophic upsurge in the reservoir(Alvarez et al. 1986; Vilarrubia and Cova 1993).

Methods

Habitat.-Field sampling was conducted fromOctober 1993 through July 1994 in the northernsection of the reservoir (from OT20'N to 07°50'N

and from 63°40'W to 63°05'W; Figure 1). We re-stricted sampling to littoral areas containing flood-ed timber, savanna, or sand and gravel beaches.Temperature (OC) , depth (m) , and substrate type(sand, gravel, flooded savanna, flooded forest)were recorded before sampling fishes. Dissolvedoxygen (mg/L) , hardness (mg/L as CaCO3), pH,and alkalinity (mg/L as HCO3- + CO3-2) weremeasured by using a Hach kit.

Fish community composition.-We sampled fish-es with 3-m and 6.5-m seines (6.3-mm mesh) toevaluate the prey fish assemblage. All specimensfrom seine hauls were preserved for taxonomicidentification and calculation of species relativeabundances. Species composition was comparedbetween seasons (rainy versus dry) and amongcapture sites (eight locations) with a chi squarecontingency table (SYSTAT 1992). Sport fishers'catches, particularly Cichla species, were exam-ined opportunistically, more or less at random withrespect to time and location, in order to increasesamples for diet and condition analyses.

To evaluate change in the piscivore assemblage,monofilament nylon gill nets (each net had three

277PISCIVORES IN GURI RESERVOIR, VENEZUELA

ings (e.g., family). The number of stomachs ex-amined, number of stomachs with food, total foodvolume, and the volumetric percentage of eachfood category by predator species was determinedfor each piscivore. Diet breadth was indexed withLevins' (1968) standardized index of niche width:B= 1/!.pij2, where Pij is the proportion of resourcei used by the consumer. MacArthur and Levins'(1967) index was used to compare diet overlapbetween all pairwise combinations of the four pis-civores: cl>jk = !.(PijPik)/IPij2, where Pij and Pik arethe proportions of resource i used by species j andspecies k, respectively, for all resources. Preylength distributions were plotted for each pisci-vore, and sizes of prey consumed by the four pis-civores were compared with (-tests and analysis ofcovariance (ANCOV A), the latter method testingfor an effect of predator size.

Results

Habitat

Our measurements of chemical and physicalvariables were consistent with the classic blackwater attributes previously described for Guri Res-ervoir (Alvarez et al. 1986; Weibezahn 1994). Weobtained low values for pH (6), hardness (0-17mg/L), and alkalinity (4-17 mg/L). Weibezahn(1994) reported pH values from 4.8 to 6.7, hard-ness from 1.4 to 4.4 mg/L, and alkalinity from 22to 78 mg/L during 1991-1992 from a station inthe lake's northern sector. Alvarez et al. (1986)reported pH from 4.8 to 6.9, hardness from 2.3 to27.0 mg/L, and alkalinity from 2.0 to 20.8 mg/Lduring the final stage of reservoir filling. Surfacewater temperatures fluctuated little (27-30°C), anddissolved oxygen was moderate to high (range 3-9 mg/L) throughout our 10-month study. The mostabundant substrates at sample sites were woodydebris and flooded terrestrial vegetation, with fewareas having sand, gravel, or rock. These latterhabitats were often associated with shorelinesmore exposed to wave action, areas that yieldedfew fishes in seine and gill-net samples.

Fish Community Composition

Seining yielded 16,642 specimens representing18 families and 50 species, most of which (90%)were from six species: Hemigrammus micropterus(Characidae) 80.2%, Curimatella immaculata(Curimatidae) 3.5%, Bryconops caudomaculatus(Charcidae) 2.1 %, Hemigrammus rodwayi (Char-acidae) 1.6%, Hemiodopsis gracilis (Hemiodonti-dae) 1.4%, and Mesonauta festivus (Cichlidae)1.2%. Species composition from seine samples did

25 x 2-m panels with one panel each 2.5-, 5.1-,and 10.3-mm-bar mesh) and multifilament nylongill nets (two nets, 63 X 2 m with 15-mm stretchmesh and one net, 63 X 2 mm, with 11-mm stretchmesh) were fished during both day and night inthe littoral zone. Gill nets were lifted after ap-proximately 6-12 h increments to reduce the num-ber of specimens collected with empty stomachs.The multifilament gill nets were identical to thoseused in the 1987 survey, and both types of gillnets were set in seven locations, which enabledcomparisons between current CPUEs and thosepreviously obtained by the EDELCA Ecology Labduring 1987 (Figure 1; EDELCA, unpublisheddata). During 1987, an average of 541 m of ex-perimental gill net was set for an average of 23 hover 12 monthly outings (mean meter-hours/outing= 12,443 :t 743 SD). During 1993-1994, an av-

erage of 160 m of experimental gill net was setfor an average of 18.5 hover 12 outings (meanmeter-hours/outing = 2,960 :t 2,084 SD). Gill netswere set within 100 m of shore, mostly in coveswith submerged timber and protection from openwind and waves. The date, starting time, sampleduration, location, and habitat features were re-corded for each gill-net set and seine haul. Gill-net catch (number and weight) was standardizedto a common unit of effort (value. 100 m-l.h-l)for each sampling period and species for the 1987and 1993-1994 surveys.

Size distributions and weight-length relation-ships.-Wet weight (nearest 0.1 kg) and totallength (TL, nearest millimeter) were recorded foreach piscivore captured by gill net or hook andline. Preserved specimens from seine hauls weremeasured to the nearest 0.1 g in the laboratory.Weight-length relationship determinations includ-ed all available piscivore specimens, regardless ofcapture method, based on the equation logW = a+ b(logTL), where W is wet weight, a is the y-intercept, TL is total length, and b is the slope ofthe relationship between W and TL. Given thatefficiency for different size-classes varies amongour sampling methods, these combined results pro-vide a better overall assessment of populationsthan results from anyone method.

Diets.-All netted specimens and a subs ampleof the angling sample were examined for stomachcontents and gonad development. Stomach con-tents were identified to the lowest possible taxo-nomic level, and the volume of each prey item wasestimated by water displacement in a graduatedcylinder. Stomach contents were divided into 17food categories based on higher taxonomic group-

')72 WILLIAMS ET AL.

TABLE 1.-Between-year comparisons of catch per unit effort (CPUE; value per 100 m of net per hour) for thedominant species and total gill-net catch between the 1985-1986 (Novoa et al. 1989), 1987 (Balbas, E1ectrificaci6n delCaronf, unpublished data), and 1993-1994 surveys at Gun Reservoir. The CPUE for individuals was not reported for1985-1986.

Prochilodus rubrotaeniatusPlagioscion squamosissimusHydrolycus scomberoidesCichla temensisCichla orinocensisOthers

Total

0.10

1.20

'The 1985-1986 data were recorded for combined Cichla species.

not vary significantly among sites (eight sites) andseasons (rainy versus dry) (X2 = 4.71; df = 10; P

> 0.91). Angling produced 181 Cichla temensis,19 C. orinocensis, and 23 fishes representing otherspecies.

Our 1993-1994 gill-net samples yielded 103specimens and 8 species. The 1987 EDELCA gill-net survey produced 428 specimens and 11 spe-cies, but four species each constituted 1 % or lessof the total weight. Based on biomass, Prochilodusrubrotaeniatus was the dominant fish in gill-netcatches from all three surveys, and Plagioscionsquamosissimus ranked second (Table 1). Total fishbiomass CPUE was essentially the same for 1987and 1993-1994, but the CPUE for number of in-dividuals captured by gill nets was 71% higherduring 1993-1994. The CPUE of P. rubrotaeniatusbased on numbers was 15% higher in the last sam-ple but biomass CPUE was 16% lower, an indi-cation of smaller average size. Overall, CPUE forgill nets was low for the piscivores (Table 1). Cich-lid numbers and biomass were very low in all threesamples, and C. orinocensis was not captured ingill nets during 1993-1994. Numeric CPUE of P.squamosissimus was higher during 1993-1994, andbiomass CPUEs of both P. squamosissimus andHydrolycus scomberoides were lower in the tworecent samples compared with the 1985-1986 data(Table 1).

was dominated by juvenile size-classes (Figure 2).Hydrolycus scomberoides was the largest of thefour piscivores (0.3-6.9 kg, 33.3-107 cm TL), fol-lowed by C. temensis (2.0-5.4 kg, 6.5-70.2 cmTL), C. orinocensis (0.3-6.9 kg, 3.4-56.5 cm TL),and P. squamosissimus (1.0-3.8 kg, 6.1-61 cmTL).

Slopes describing weight-length relationshipswere virtually the same for C. temensis, C. ori-nocensis, and P. squamosissimus (Figure 3). Hy-drolycus scomberoides revealed a shallower slopeand larger weight-length intercept than the otherthree piscivores because of its more elongate bodyform. Variance around the log-log regression mod-el was minimal for all four species, an indicationof little intraspecific variation in body condition.

Size Distributions and Weight-LengthRelationships

Based on the total specimens obtained from allthree survey methods, size frequency distributionsof the two Cichla species were similar, except thata greater proportion of juveniles constituted oursample of C. orinocensis (Figure 2). The H. scom-beroides sample was dominated by adult size-classes, whereas the sample of P. sauamosissimus

Diets

Characiform fishes (Characidae and Hemiodon-tidae) were the most common prey of C. temensis(70.7%) and P. squamosissimus (61.7%; Table 2).Cichla orinocensis consumed mostly cichlids(66.3%), and H. scomberoides consumed primarilypimelodid catfishes (54.2%), although the smallnumber of diet samples for C. orinocensis and H.scomberoides make these results tentative. Themost common characiforms identified from C. te-mensis stomachs were Hemigrammus micropterus,Bryconops caudomaculatus and Hemiodopsis grac-ilis. The cichlid Crenicichla wallacei and the char-acid H. micropterus were the most common fishesin stomachs of C. orinocensis.

The diet of C. temensis was most similar to thatof P. squamosissimus, followed by C. orinocensisand H. scomberoides (Table 3). Cichla orinocensishad low diet overlap with P. squamosissimus andnegligible overlap with H. scomberoides, whereasdiets of P. squamosissimus and H. scomberoideswere similar (Table 3). Each ni!;civnre nnmJl11tinn

PISCIVORES IN GURI RESERVOIR, VENEZUELA 279

Cichla temensisMean=35.3680=13.13

Cichla orinocensisMean=54.6880=18.91

>-uc:Q):J0-Q)'-

u.

Plagioscion squamosissimusMean=27.20SD=12.76

scomberoidesMean=54.6SD=18.91

II) 0 II) 0 II) 0 II) 0 II) 0 II) 0 II) 0 II) 0 II) 0 II) 0 II) 0, C\lC\I(')(')VVII)II)<O<OI'-I'-a>a>O)O)oo...

0 , , , , , , ., """"" 11)011)011)011)011)011)011)011)011)0" .... ... C\I C\I (') (') v v II) II) <0 <0 I'- I'- a> a> 0) II) 0 II)

0) 0 0

Total Length Interval (cm)

FIGURE 2.-Size distributions of the four most common Guri Reservoir piscivores, based on specimens captured bygill nets, seines, and angling.

slopes of prey length-predator length for C. te-mensis and P. squamosissimus were significantlydifferent, so apparent species effects could not beinterpreted.

Discussion

Habitat

The extremely low nutrient content of Guri Res-ervoir's black waters and its low retention time arethe factors that probably limit secondary produc-tion, including fishes. Gill-net CPUE estimatesfrom three surveys during the period 1985-1994indicate that the standing stock of fish biomass inGuri Reservoir is low relative to reservoirs world-wide. Total fish CPUEs in Guri during 1987 and1993-1994 (0.59 and 0.60 kg. 100 m-1.h-l) werewithin the range reported for Itaipu Reservoir, Pa-rana River basin, Brazil (0.3 to 0.65 kg.100m-l.h-l), which is characterized as unproductive(Agostinho and Zalewski 1994). The hi2hest

consumed only a small portion of the availableprey taxa, with standardized diet breadths rangingbetween 0.085 and 0.105 (Table 3).

Except for P. squamosissimus, most piscivoresconsumed fish prey between 3 and 6 cm TL (Figure4). Many of the small P. squamosissimus in oursample consumed aquatic insects and penaeidshrimp shorter than 3 cm. Mean prey sizes (log-transformed prey TL) for all pairwise combina-tions of piscivores were significantly different (t-test, P < 0.01), with the exception of the com-parison between C. orinocensis and P. squamosis-simus (P = 0.18, due to the small sample size forC. orinocensis).

Predator size and prey size were strongly andpositively correlated for each of four piscivores(Figure 5). Hydrolycus scomberoides consumedsignificantly larger prey than P. squamosissimus(ANCOVA; P < 0.01), but no other species pairsrevealed statistically significant differences in preysize after adjustment for predator length. However,

280 WILLIAMS ET AL.

1.0

-0;~

.:E01Ow$:

010

-J

-3.0

1.0PlagioscionW = - 5.34 + 3.28Lr2 = 0.99

HydrolycusW = - 4.47 + 2.64Lr2 = 0.98 /

0

-0; 0.0:=:.

:E0>"ii) -1.0~0>0-J -2.0

/ n = 56 n = 19-3.0 .. -.~

0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0

Log Total Length (cm) Log Total Length (cm)

FIGURE 3.-Weight (W) to length (L) relationships for C. temensis, C. orinocensis, ,Po squamosissimus, and H.scomberoides. Data points with equal values on both axes appear as a single dot.

CPUE value for Guri Reservoir is from a 1985-1986 survey (Novoa et al. 1989), which was con-ducted in newly flooded savanna coves near Man-teco (Figure 1). Fish biomass appears to have de-clined just 2 years after impoundment in Gun Res-ervoir, compared with 6 years after impoundmentin Itaipu Reservoir.

TABLE 2.-Data on fish stomachs and volumetric per-centages of major food categories consumed by four GuriReservoir piscivores.

Total volume (nIL)Number of stomachsNumber of stomachs

with foodFood category (%)

PlantsCrustaceaInsectsUnidentified fishCharacidaeHemiodontidaePimelodidaeSciaenidaeCichlidaeSynbranchidaeTadpoles

48.317

48.8

5432.516

Fish Community Composition

The domination of the littoral zone in Guri Res-ervoir by a few small fish species is not unusualcompared with Itaipu Reservoir, where 90% of thefishes comprised only nine species, mostly char-aciforms (Agostinho and Zalewski 1994). The nu-merically dominant species in our seine samples,H. micropterus. feeds on zooplankton, seeds, andaquatic and terrestrial invertebrates in shallownearshore waters. Species of Hemigrammus fromVenezuela's savanna region have an opportunisticlife history strategy that enables rapid colonizationof disturbed habitats (Winemiller 1989), and Guri

58 10 36 6

0.20.102.3

60.010.700

21.25.50

00.800

17.414.500

66.301.0

2.05.13.87.4

56.84.9

19.30.6000

0001.5

10.20

54.234.2000

0

-1

-2

.0

.0

.0

PISCIVORES IN OURI RESERVOIR. VENEZUELA 281

TABLE 3.-Standardized diet breadths and pairwise dietoverlaps for the four most common piscivores at Gun Res-ervoir.

Diet overlap for:

Plagio- Hydro-Cichla Cichla scion lycus

Diet temen- orino- squamo- scrom-breadth sis censis sissimus boidesSpecies

C. lemensisC. orinocensisP. squamosissimusH. scomberoides

0.0860.0640.1050.085

0.571

0.880.24

1

0.140.03

0.41

0 2 4 6 8 10

Mean Prey Length (cm)

FIGURE 4.-Mean prey lengths (whiskers = :!:SE) forfour cornrnon Gun Reservoir piscivores: C. temensis, C.orinocensis, P. squamosissimus, and H. scomberoides.

Gill-net CPUE data suggest that abundance ofthe two Cichla species in the littoral zone of Gunhas decreased only slightly from 1987 to 1994,whereas P. squamosissimus has increased slightly.Novoa et al. (1989) reported declines in the per-centages of Cichla species in gill-net samples overthree consecutive years (from 6.6% to 4.9% to4.5% between 1985-1987), whereas P. squamo-sissimus averaged 10.8% of their gill-net catchover this period. According to our 1993-1994seine samples, P. squamosissimus appeared to havegood recruitment at least to 40 cm TL. Because ithas pelagic eggs and larvae, P. squamosissimus isless impacted by seasonal water level fluctuationsthan Cichla spp., which construct and guard nestsin shallow littoral regions (Zaret 1980).

Reservoir's annual water level fluctuations createa changing environment to which H. micropterusis well suited. No prior data exist for relative abun-dances of species in seine samples, but our overallspecies list contains 46 of the 68 species collectedby EDELCA during 1985-1987 and reported byLasso et al. (1989). In addition, our survey pro-duced eight species not reported by Lasso et al.(1989). Among these was the dominant species,H. micropterus, and H. rodwayi, the fourth mostabundant species from our seine samples. The dif-ferences in community samples between the twosurvey periods may reflect temporal change in thefish community, differences in the locations sur-veyed (the earlier survey included the lake's south-ern region), or a combination of both. It is unlikelythat the absence of the abundant H. micropterusand H. rodwayi from the earlier survey was dueto sampling error.

Prochilodus rubrotaeniatus dominated the bio-mass in all gill-net surveys of Guri Reservoir (No-voa et al. 1989), and there are several reasons whythis large migratory characiform continues tothrive in the lake. Prochilodus rubrotaeniatus is adetritivore, and as such, probably is less resource-limited than other species that are more directlydependent on primary production. Detritivores, es-pecially prochilodontids, also constitute a majorfraction of the fish biomass of the Rio Negro, RioOrinoco, and Itaipu Reservoir (Goulding et al.1988; Novoa 1989; Agostinho and Zalewski1994). Prochilodus rubrotaeniatus apparently mi-grate up the major tributaries of Guri Reservoirfor spawning, and their larval and juvenile stagesinhabit floodplains in the lake's southern region.We captured only two juveniles within the lake'snorthern region. Spawning migrations to upstreamriver floodplains and juvenile use of floodplainhabitats have been described for Prochilodus scro-fa in the Itaipu Reservoir of southern Brazil (Agos-tinho et al. 1993).

Size Distributions and Weight-LengthRelationships

Fishes from our survey showed a greater in-crease in mass with increasing body length thanthose in an earlier study (Novoa et al. 1989), whichimplies better condition as a result of greater percapita prey availability. Condition slopes (basedon log-transformed weight-length data) werehigher for all four piscivores in our study than ina previous study in Gun Reservoir (Novoa et al.1989): 2.77 versus 3.14 for Cichla spp. combined;2.52 versus 2.64 for P. squamosissimus; and 2.52versus 3.27 for H. scomberoides. This pattern sug-gests that a recent decline in densities of Cichlaspp. from fishing or predation mortality might haveresulted in 2rowth compensation throu!!h !!reater

282 WILLIAMS ET AL.

2.0

1.5

1.0

0.5

0.0

-0.5

-1.0

Cichla temensisY = 0.52X - 0.14r2 = 0.71 ~ %

coB ~ 00_ .::.~-~ o-~~:~;;~~~:~:: m ~O

00 o 00 =

o

E.£01.Q.I::0>cQ)-.J

70+0f->-Q)It n = 141

2.0Plagioscion squamosissimusY = 1.52X - 1.87r2 = 0.68

Hydrolycus scomberoidesY = 1.42X - 1.66r2 = 0.56

00 00

:;1:'0 0ftft 0

0-/00°8°m ... 0

GO0 00

0

E0 1.5"C)0- 10.I:: .OJ~ 0.5-.J

"§ 0.00f-

>- -0.5~~ n = 56 n = 12- -1.0

0.60.8 1.0 1.2 1.4 1.6 1.82.0 0.60.8 1.0 1.2 1.4 1.6 1.82.0

Predator Total Length log(cm) Predator Total Length log(cm)

FIGURE 5.-Relationships of prey length (Y) to predator length (X) for four common Guri Reservoir piscivores. Datapoints with equal values on both axes appear as a single dot.

resource availability in this hyperoligotrophic eco-system.

wallacei, other cichlids, and the abundant characidH. micropterus are common. In contrast, C. te-mensis (especially larger individuals) apparentlyforage in deeper and less structured habitats wherethey capture midwater characiforms, such as Bry-conops caudomaculatus and Hemiodopsis gracilis.

The diets of C. temensis and P. squamosissimuswere similar based on prey type, but C. temensisconsumed larger prey. (However, relatively fewlarge P. squamosissimus were captured from shal-low littoral habitats.) Competition between Cichlaspecies and H. scomberoides was unlikely as thesetwo species consumed similar-sized prey but hadvery low diet similarity. Hydrolycus scomberoidesshowed the same trends with C. orinocensis. Cichlaorinocensis and P. squamosissimus consumed dif-ferent sizes and types of prey. Plagioscion squa-mosissimus and H. scomberoides consumed sig-nificantly different-sized prey, yet did show someoverlap based on prey type. Whereas P. squamo-sissimus ate mostly characids and pimelodids, H.

Diets

On average, H. scomberoides consumed the larg-est prey, and P. squamosissimus consumed thesmallest prey. This result may reflect resource par-titioning or it may be an artifact of sampling tech-niques that selected for larger H. scomberoides.The two Cichla species did not differ in the meansize of prey consumed, even when adjusted forpredator length. The diets of the two cichlids weresimilar based on prey type as well, so there waspotential for resource competition. Even so, C. te-mensis consumed mostly characids, followed bycichlids, whereas C. orinocensis consumed mostlycichlids, followed by characids. Gil et at. (1993)and Novoa (1993) reported similar findings fromGun Reservoir for C. temensis and C. orinocensis.Most C. orinocensis probably forage in shallow,structured habitats near shore where Crenicichla

PISCIVORES IN GURI RESERVOIR. VENEZUELA 283

scomberoides ate mostly pimelodids and, to a less-er extent, sciaenids (i.e., P. squamosissimus). Hy-drolycus scomberoides therefore has the potentialto influence abundance of P. squamosissimus.

Diet breadths for all four piscivores were low:P. squamosissimus had the broadest diet, C. ori-nocensis had the narrowest diet (but sample sizeswere small), and C. temensis and H. scomberoideshad intermediate diet breadths. Cichla temensis fedheavily on the three most abundant fishes in thelittoral zone of the reservoir, an indication of anopportunistic feeding strategy. Similarly, C. ori-nocensis consumed the two most abundant littoralzone fishes, but not in accordance with their rankabundance. Plagioscion squamosissimus consumedmostly small characids and pimelodid catfishes,the latter being bottom-dwellers most abundant indeeper waters farther offshore. Despite the factthat its superior mouth, large pectoral fins, andkeeled chest are designed for midwater and surfacefeeding, H. scomberoides in Guri Reservoir fedheavily on benthic catfishes and P. squamosissi-mus.

that average fish size has decreased. The two prin-cipal sport fishes, Cichla spp., appear to have de-clined slightly in numbers and biomass in asso-ciation with the waning of the lake's trophic up-surge phase. To sustain the sport fishery for Cichlaspp. in Guri, current harvest limits should be en-forced. Given the exceedingly low primary andsecondary productivity of the reservoir, a policyof catch-and-release sportfishing would better pro-tect fragile stocks of Cichla spp. Yet, regulationof sportfishing is only one aspect of fisheries man-agement in Guri Reservoir, as commercial fishingis being considered. The initial proposal for com-mercial fishing in Guri Reservoir was based uponmodel projections that used yearly fish productionestimates (80-220 kg/ha) extrapolated from ablock-net sample from a single cove (flooded sa-vanna) in the middle region of the reservoir duringthe trophic upsurge phase (Novoa et al. 1989). Themodel simulations indicated that even moderatesportfishing negatively impacted populations ofCichla spp. and that commercial fishing reducedthe impact of sportfishing on Cichla spp. becausemany large predators that feed on small Cichla spp.would be removed by the large-mesh gill nets usedto harvest large Prochilodus rubrotaeniatus. Giventhe model's gross assumptions regarding the influ-ence of food web structure on population dynam-ics, we urge caution in the interpretation of thesesimulations. Also, the reservoir's heterogeneity,especially gradients of primary productivity,should be considered carefully when evaluatingalternatives for fishcries management on a whole-lake basis (Silver et al. 1986). To improve futuremodel projections, additional surveys that com-pare sites along the reservoir's south-north (i.e.,intlow-dam) productivity gradient should be con-ducted.

AcknowledgmentsThis study could not have been done without

support and assistance from the staff of the GuriEcology Laboratory (CVG-EDELCA): Jose Mar-tin Lubin, Asterio Farreras, Carlos Ron, TeopistaSuarez de Aponte, John Stredel, Rafael Suarez,Jose Llerandi, Miguel Morales Miranda, and Car-los Nava. We sincerely thank all of them. We alsothank Kate Ryan Williams, Jeanie Gessner, DavidJepsen, and Tamara McGuire for assisting withsome of the field surveys. We thank Raymond Car-roll for helpful statistical advice and Steve Mir-anda for valuable comments that helped us im-prove the manuscript; however, we alone acceptresponsibilitv for data intemretations and man-

Management ImplicationsSport fish catch rates often are inversely cor-

related with reservoir age (Ploskey 1986). Becauseof the low nutrient content of the region's soilsand Guri Reservoir's low retention time, the initialpulse of nutrients associated with reservoir fillingwas probably weak and short lived. Alvarez et al.(1986) estimated yearly fish production for Guriat just 10 kg/ha, based on the morphoedaphic index(conductivity/lake depth) and biological assump-tions borrowed from work on Lake Kariba in Af-rica (Marshall et al. 1982). For comparison, theycited the yearly estimate of 23.5 kg/ha for Bro-kopondo Reservoir in Suriname (Kapetsky 1978),another unproductive reservoir of the GuyanaShield. They also cautioned against the use of es-timates of production based on the trophic upsurgephase of reservoir development. The gill-net dataof Novoa et al. (1989) for the period 1985-1986(Table 1) were collected from some locations dif-ferent than ours, so the difference between theirtotal hourly CPUE (1.2 kg/lID m) and our valuesfor 1987 and 1993-1994 (0.59 and 0.60 kg/I 00 m)probably reflects the combined influence of a de-cline in the lake's trophic upsurge and within-lakespatial heterogeneity of productivity (nets werefished day and night in both studies).

Our gill-net CPUE data indicate that the lake'salready low overall fish abundance has notchanged appreciably between 1987 and 1994 but

284 WILLIAMS ET A

agement recommendations. Financial support wasprovided in part by the Fundaci6n Fluvial de LosLlanos, Texas A&M University's Jordan Institute(International Loan) and International EducationFee Scholarship.

80's. American Fisheries Society, Southern Divi-sion, Reservoir Committee, Bethesda, Maryland.

Kapetsky, J. 1978. The Brokopondo Reservoir: Fisheryyield potential, fishery research and fishery devel-opment. FAO (Food and Agriculture Organizationof the United Nations), Technical Report, Rome.

Kimmel, B. L., and W. Groeger. 1986. Limnologicaland ecological changes associated with reservoiraging. Pages 103:-109 in Hall and Van Den Avyle( 1986).

Lasso, C. A., A. Amigos, D. Novoa, and F. Ramos. 1989.La ictiofauna dellago de Guri: composicion, abun-dancia y potencial pesquero. Parte I: consideracionsgenerales e invetario de la ictiofauna dellage Guricon breve description de las especies de interes parala pesca deportiva y comercial. Sociedad de Cien-cias Naturales La Salle 49:141-158.

Levins, R. 1968. Evolution in changing environments:some theoretical explorations. Princeton UniversityPress, Princeton, New Jersey.

Lowe-McConnell, R. H. 1973. Summary: reservoirs inrelation to man-fisheries. Pages 641-654 in W. C.Ackermann, G. F. White, E. B. Worthington and J.L. Ivens, editors. Man-made lakes: their problemsand environmental effects. American GeophysicalUnion, Washington, D. C.

MacArthur, R. H., and R. Levins. 1967. The limitingsimilarity, convergence, and divergence of coexist-ing species. American Naturalist 101:377-385.

Marshall, B. E., F. Junor, and L. Langerman. 1982. Fish-eries and fish production on the Zimbabwean sideof lake Kariba. Kariba Studies: National Museumand Monuments of Zimbabwe 10:175-231.

Morales, L. C., and S. Gorzula. 1986. The interrelationsof the Caronf River basin ecosystems and hydro-electric power projects. Interciencia II :272-277.

Noble, R. L. 1986. Management of reservoir fish com-munities by influencing species interactions: pred-ator-prey interactions in reservoir communties.Pages 137-143 in Hall and Van Den Avyle (1986).

Novoa, D. F. 1989. The multispecies fisheries of theOrinoco river: development, present status, andmanagement strategies. Canadian Special Publica-tion of Fisheries and Aquatic Sciences 106.

Novoa, D. E 1993. Aspectos generales sobre la biol-ogfa, pesqueria, manejo y cultivo del pav6n (Cichlaorinocensis y C. temensis) en ellago de Guri y otrasareas de la regi6n Guayana. Natura 96:34-39 (So-ciedad de Ciencias Naturales La Salle, Caracas,Venezuela.)

Novoa, D. E, J. Koonce, A. Locci, and F. Ramos. 1989.La ictiofauna dellago de Guri: composicion, abun-dancia y potencial pesquero. II. Evaluacion del po-tencial pesquero dellage Guri y estrategias de or-denamiento pesquero. Memoria de la Sociedad deCiencias Naturales La Salle 49:159-197.

O'Brien, W. J. 1990. Perspectives on fish in reservoirlimnology. Pages 209-225 in K. W. Thorton, B. L.Kimmel and E E. Payne, editors. Reservior lim-nology: ecological perspectives. Wiley, New York.

Ploskey, G. R. 1986. Management of the physical and

References

Agostinho, A. A., A. E. A. de M. Vazzoler, L. C. Gomes,and E. K. Okada. 1993. Estratificaci6n espacial ycomportamiento de Prochilodus scrota en distintasfases del cicio de vida, en la planicie de inundaci6ndel alto rio Parana y embalse de Itaipu, Parana,Brasil. Revista de Hydrobiologia Tropical 26:79-90.

Agostinho, A. A., and M. Zalewski. 1994. The depen-dence of fish community structure and dynamics onfloodplain and riparian ecotone zone in Parana Riv-er, Brazil. Hydrobiologia 303:141-148.

Allan, J. D., and A. S. Flecker. 1993. Biodiversity con-servation in running waters. BioScience 43:32-43.

Alvarez, E., L. Balbas, I. Massa, and J. Pacheco. 1986.Aspectos ecologicos del embalse Guri. IntercienciaII :325-333.

Balbas, L., and J. Pacheco. 1989. Los peces del embalseGuri y del tramo bajo del Rio Caroni, incluyendoalgunas consideraciones ecol6gicas acerca de la in-fluencia de la formaci6n del embalse sobre la ictio-fauna de la cuenca. Corporaci6n Venezolana de Gui-ana-Electrificaci6n del Caroni, Report, Guri, Ven-ezuela.

Barthem, R. B., M. Petrere, and M. C. Lambert. 1991.Life strategies of some long-distance migratory cat-fish in relation to hydroelectric dams in the Amazonbasin. Biological Conservation 55:339-345.

Feamside, P. M. 1989. Brazil's Balbina dam: environ-ment versus the legacy of the pharaohs in Ama-zonia. Environmental Management 13:401-423.

Gil, C. E., J. Andrada, E. Mendez, and J. M. Salazar.1993. Estudio preliminar sobre alimentaci6n encautiverio y contenido estomacal de Cichla temensisdel embalse Guri, Estado Bolivar, Venezuela. Na-tura 96:42-47. (Sociedad de Ciencias Naturales LaSalle, Caracas, Venezuela.)

Goddard, J. A., and L. C. Redmond. 1986. Stocktonlake: prolonging the "boom," managing a new largereservoir with minimum length limits. Pages 203-210 in Hall and Van Den Avyle (1986).

Gonzalez, E., J. Paolini, and A. Infante. 1989. Waterchemistry, physical features and primary productionof phytoplankton in a tropical blackwater reservoir(Embalse de Guri, Venezuela). Intemationale Ver-einigung fUr Theoretische und Angewandte Lim-nologie Verhandlungen 24:1477-1481.

Goulding, M., M. L. Carvalho, and E. G. Ferreira. 1988.Rio Negro, rich life in poor water: Amazonian di-versity and foodchain ecology as seen through fishcommunities'. SPB Academic Publishing, TheHague, Netherlands.

Hall, G. E., and M. J. Van Den Avyle, editors. 1986.Reservoir fisheries management: strategies for the

285PISCIVORES IN GURI RESERVOIR, VENEZUELA

Pages 199-213 in A. D. Hasler, editor. Coupling ofland and water systems. Springer Verlag, New York.

SYSTAT. 1992. SYSTAT for windows, version 5 edi-tion. SYSTAT, Evanston, Illinois.

Vilarrubia, T. V., and M. Cova. 1993. Estudio sobre ladistribuci6n y ecologfa de macrofitos aquaticos enel embalse de Guri. lnterciencia 18:77-82.

Weibezahn, F. H. 1994. Lake Guri (Venezuela): prelim-inary limnological characterization of a large trop-ical blackwater reservoir. lnternationale Revue derGesamten Hydrobiologie 79:47-60.

Winemiller, K. O. 1989. Patterns of variation in lifehistory among South American fishes in seasonalenvironments. Oecologia 81:225-241.

Winston, M. R., C. M. Taylor, and J. Pigg. 1991. Up-stream extirpation of four minnow species due todamming of a prairie stream. Transactions of theAmerican Fisheries Society 120:98-105.

Zaret, T. M. 1980. Life history and growth of Cichlaocellaris, a predatory South American cichlid. Bio-tropica 12:144-157.

Received May 28, 1996Accepted October 21. 1997

chemical environment: effects of water. Pages 86-97 in Hall and Van Den Avyle (1986).

Prosser, N. S. 1986. An overview of reservoir fisheriesproblems and opportunities resulting from hydro-power. Pages 238-246 in Hall and Van Den Avyle( 1986).

Randall, R. G., J. R. M. Kelso, and C. K. Minns. 1995.Fish production in freshwaters: are rivers more pro-ductive than lakes? Canadian Journal of Fisheriesand Aquatic Sciences 52:631-643.

Rzoska, J. 1966. The biology of reservoirs in the U. S.S. R. Pages 149-154 in R. H. Lowe-McConnell,editor. Man-made lakes. Institute of Biology Sym-posia 15.

Schwassman, H. O. 1978. Times of annual spawningand reproductive strategies in Amazonian fishes.Pages 187-200 in J. E. Thorpe, editor. Rhythmicactivity of fishes. Academic Press, London.

Silver, J. R., W. J. Foris, and M. C. McInerny. 1986.Spatial heterogeneity in fish parameters within res-ervoirs. Pages 122-136 in Hall and Van Den Avyle( 1986).

Sioli, H. 1975. Amazon tributaries and draina~e basins.