Structure of fish assemblages along environmental gradients in floodplain lakes of the Orinoco River.
Key words: abiotic factors; community organization; multivariate analyses; Orinoco River floodplain; piscivory; predator-prey interactions; predictability; seasonality; tropical ecology; turbidity; water; Venezuela; vision.INTRODUCTION
The processes that control structure of freshwater fish assemblages have been clarified by the use of simple conceptual models (Schlosser 1987, Robinson and Tonn 1989, Tonn et al. 1990, Rahel and Hubert 1991, Jackson et al. 1992). Most of these models explicitly relate assemblage structure to the abiotic environment, species interactions, and biogeographical events (e.g., Schlosser 1987, Robinson and Tonn 1989, Tonn et al. 1990, Rahel and Hubert 1991, Kodric-Brown and Brown 1993). The generality of the relationships between assemblage structure and the environment has been tested across distinct assemblages over several geographical regions (Tonn et al. 1990). Comparative surveys of lakes have provided excellent opportunities for examining the effects of biotic and abiotic factors on fish assemblage structure at large and intermediate geographical scales that encompass very substantial variation in the abiotic environment (Jackson and Harvey 1989, Tonn et al. 1990, Hinch 1991). Despite the essentially correlative nature of comparative surveys, appropriate designs incorporating repeated sampling support inferences about ecological process (Tonn and Magnuson 1982). Conclusions from such surveys have sometimes been supplemented with small-scale experiments involving manipulation of piscivore abundances (Tonn et al. 1992, Hinch et al. 1994).
Robinson and Tonn (1989) postulated that piscivory, in conjunction with a small number of environmental variables including lake maximum depth, surface area, and isolation, have widespread influence on the structure of fish assemblages in small temperate lakes. This simple conceptual model is consistent with the results of more recent studies (Tonn et al. 1990, Hinch 1991, Hinch et al. 1991). All tests of Robinson and Tonn's model, however, have been conducted in temperate regions. The marked taxonomic and environmental contrasts between temperate and tropical lakes make the latter a very suitable system for testing the robustness of Robinson and Tonn's model. Tropical floodplain lakes are comparable in size to the temperate lakes for which fish assemblages have been most intensively analyzed. Tropical floodplain lakes support highly diverse fish assemblages that differ taxonomically from their temperate counterparts at the ordinal level (Lowe-McConnell 1975, Welcomme 1979). Abiotic conditions in floodplain lakes (e.g., temperature, seasonal flooding) also differ considerably from those in temperate lakes (Lowe-McConnell 1975, Welcomme 1979). Over a broad geographic range in South America, from Venezuela to Argentina, a number of widespread fish genera recur as dominant taxa in floodplain lake assemblages, yet assemblage composition can vary strikingly even among neighboring lakes (Bonetto et al. 1970a, b, Mago-Leccia 1970, Goulding et al. 1988, Rodriguez and Lewis 1994). Major reviews of tropical river systems (Lowe-McConnell 1975, 1987, Welcomme 1979, 1985, Goulding et al. 1988) fail to show clear relationships between characteristics of floodplain lakes and the structure of their fish assemblages. Although Welcomme's (1979, 1985) summaries point out that size and bottom type of lakes may influence fish species composition and body size, several authors have suggested instead that fish assemblages in these lakes are largely random associations of species (Bonetto et al. 1970a, b, Lowe-McConnell 1987: 154, 274, 304, Goulding et al. 1988:106-107, 116).
We have tested the applicability of Robinson and Tonn's model to floodplain lakes of the Orinoco River, Venezuela. Our test involves the analysis of fish assemblages in 20 lakes of three different regions in relation to environmental variables and piscivory. The analysis is based on abundances of species as shown by standardized sampling of fishes in the early and late dry seasons of two consecutive years.
SITE DESCRIPTION AND METHODS
Study area
The annual cycle in Orinoco floodplain lakes can be divided into a high-water season, when the lakes are connected to each other, and a dry season, when lakes are isolated. Each season lasts [approximately equals] 6 mo. Floodwater usually enters the lakes in June, after which the water level rises rapidly; in 2 mo the depths of lakes can increase from 1-2 m to [is greater than] 7 In. During the high-water season, connections between the river and the lakes facilitate movement of adult and juvenile fish as well as dispersal of eggs and larvae, which appear predominantly during this period (Lowe-McConnell 1975, 1987, Welcomme 1979, 1985). Population densities of all fish species are low in the lakes until December, when the flood recedes and lakes once more become isolated and spatially discrete (Rodriguez and Lewis 1994). During isolation, the depth of lakes declines because of evaporation, and transparency decreases as inorganic particulate matter is resuspended by wind and phytoplankton populations increase (Hamilton and Lewis 1987). Under these conditions, the water is often turbid and brown; Secchi transparency may be [is less than] 5 cm. Conductance increases during isolation as evaporation increases the concentration of major ions (Hamilton and Lewis 1987). Species richness of fishes is high during isolation, averaging 35 species per lake (Rodriguez and Lewis 1990).
Fish and environmental data were collected from 20 floodplain lakes (median surface area = 16 ha; range = 3-251 ha) in three regions of the lower Orinoco River basin in Venezuela (Table 1): near Caicara (4 lakes); near Ciudad Bolivar, [approximately equals] 400 km downstream from Caicara (12 lakes); and along the Caura River, near its confluence with the Orinoco, between Caicara and Ciudad Bolivar (4 lakes). Every year the Caura lakes are flooded by transparent and ionically dilute blackwater from the Caura River, whereas the Caicara and Ciudad Bolivar lakes are flooded by more turbid water of higher ionic concentration (whitewater) from the Orinoco River. Human population densities are low in the study area; the river and adjacent waterbodies are in nearly pristine condition.
TABLE 1. Geographic location and sample sizes for fish collection sites in the Orinoco floodplain.
Lake Region
a Aguilera Ciudad Bolivar
b Bartolico Caicara
c Burro Caicara
d Cano Largo Ciudad Bolivar
e Castillero Caicara
f Curita Ciudad Bolivar
g Jobera Ciudad Bolivar
h Larga II Caicara
i Madera Caura
j Maldonado Ciudad Bolivar
k Merecure Ciudad Bolivar
l Orsinera Ciudad Bolivar
m Paramuto Caura
n Paso Acosta Caura
o Playa Blanca Ciudad Bolivar
p Potrero Caura
q Tejita Ciudad Bolivar
r Terecaya Ciudad Bolivar
s Tineo Ciudad Bolivar
t Venado Ciudad Bolivar
Coordinates
Lake North West
a Aguilera 08 [degrees] 11'30" 63 [degrees] 26'45"
b Bartolico 07 [degrees] 38'30" 66 [degrees] 07'00"
c Burro 07 [degrees] 39'30" 66 [degrees] 12'30"
d Cano Largo 08 [degrees] 14'50" 63 [degrees] 15'55"
e Castillero 07 [degrees] 38'20" 66 [degrees] 09'00"
f Curita 08 [degrees] 09'30" 63 [degrees] 35'55"
g Jobera 08 [degrees] 09'55" 63 [degrees] 32'05"
h Larga Il 07 [degrees] 38'55" 66 [degrees] 13'40"
i Madera 07 [degrees] 33'40" 65 [degrees] 00'50"
j Maldonado 08 [degrees] 06'00" 63 [degrees] 45'50"
k Merecure 08 [degrees] 12'55" 63 [degrees] 17'25"
1 Orsinera 08 [degrees] 09'25" 63 [degrees] 34'25"
m Paramuto 07 [degrees] 33'05" 64 [degrees] 59'50"
n Paso Acosta 07 [degrees] 21'55" 65 [degrees] 11'40"
o Playa Blanca 08 [degrees] 09'10" 63 [degrees] 35'00"
p Potrero 07 [degrees] 36'40" 64 [degrees] 49'50"
q Tejita 08 [degrees] 09'15" 63 [degrees] 26'55"
r Terecaya 08 [degrees] 11'30" 63 [degrees] 27'20"
s Tineo 08 [degrees] 11'25" 63 [degrees] 28'20"
t Venado 08 [degrees] 10'30" 63 [degrees] 37'35"
Catch numbers
Lake Total Tetras
a Aguilera 577 7
b Bartolico 471 51
c Burro 127 3
d Cano Largo 1514 978
e Castillero 761 82
f Curita 172 1
g Jobera 1248 626
h Larga Il 1838 164
i Madera 142 5
j Maldonado 1243 628
k Merecure 1941 596
1 Orsinera 1680 830
m Paramuto 199 31
n Paso Acosta 1518 749
o Playa Blanca 1766 797
p Potrero 775 248
q Tejita 518 346
r Terecaya 1271 2
s Tineo 764 216
t Venado 344 19
Fish sampling and taxonomy
The 20 lakes were sampled over a 20-d period in January (early dry season) and in April or May (late dry season) during 1987 and 1988. One lake (Lake Madera) was not sampled in the late dry season of 1988. Fish samples were collected with an aluminum electrofishing boat following the guidelines in Novotny and Priegel (1974) and Reynolds (1983). Current intensity (fast-rise, slow-decay, 4-ms pulses at 120 Hz) usually was maintained within the 3- to 6-amp range recommended by Reynolds (1983). Fish samples were collected during the day (0700-1815) along shoreline intervals selected at random. Between two and seven intervals, summing to 50-280% of a lake's perimeter, were sampled during each survey. Individual sampling runs lasted 20 min and covered [approximately equals] 900 m of shoreline.
Proper analysis and interpretation of assemblage patterns require a sampling method that provides an unbiased representation of the actual assemblages. Electrofishing with pulsed current is one of the least selective of all active fishing methods (Lagler 1978). In the present study, it yielded abundant catches (a total of 18,869 fish), a very similar distribution of species among orders, but a higher mean number of species per lake than that obtained for five other South American lake systems sampled by use of rotenone or of a variety of net types (Rodriguez and Lewis 1990). Although electrofishing can be biased against small fish (Reynolds 1983, Zalewski and Cowx 1990), many small species were captured in great abundance and the distribution of species sizes (standard lengths, SL) showed little indication of a distinct lower cutoff (Table 2). Species inadequately sampled by electrofishing during the day include those living in close association with floating vegetation (e.g., Apistogramma spp., Polycentrus schomburgkii, Synbranchus marmoratus; M. A. Rodriguez, personal observation), and small catfishes that hide in logs and other submerged substrata (Lowe-McConnell 1975, Rodriguez et al. 1990). These biases should not greatly affect the results, however, because few takes had abundant floating vegetation (Table 3), and all samples were collected in daylight, so that sampling biases related to diurnal rhythms would be similar across lakes.
[TABULAR DATA 2 NOT REPRODUCIBLE IN ASCII]
TABLE 3. Distribution parameters (medians, lower quartiles (LQ), upper quartiles (UQ), minima, and maxima) for environmental variables in Orinoco floodplain lakes. Within- and supra-lake variables without explicit mention of units are expressed as proportions.
Median LQ UQ
Within-lake variables
Nearshore depth (m) 0.80 0.65 1.15
Submerged vegetation 0.14 0.01 0.53
Overhanging vegetation 0.12 0.02 0.38
Floating grasses 0.01 0.00 0.21
Open shoreline 0.15 0.03 0.55
Submerged deadwood 0.02 0.00 0.08
Water hyacinth 0.00 0.00 0.02
Other floating plants 0.00 0.00 0.00
Emergent rocks 0.00 0.00 0.00
Whole-lake variables
Secchi transparency (cm) 22.5 10.0 40.0
Conductance ([micro]S/cm) 30.5 13.0 46.5
Oxygen concentration (mg/L) 7.0 5.9 7.9
pH 6 6 6
Surface area (ha) 16.5 9.5 113.9
Fetch (km) 0.89 0.58 1.77
Breadth (km) 0.31 0.17 0.82
Perimeter (km) 2.52 1.39 6.18
Shoreline development 1.64 1.41 1.91
Maximum depth (m) 1.50 1.05 2.35
Supra-lake variables
Distance from river channel (km) 0.8 0.4 1.4
Distance from floodplain edge (km) 0.6 0.3 1.0
Non-forested perimeter 0.05 0.00 0.40
Min. Max.
Within-lake variables
Nearshore depth (m) 0.45 1.55
Submerged vegetation 0.00 0.97
Overhanging vegetation 0.00 1.00
Floating grasses 0.00 0.99
Open shoreline 0.00 1.00
Submerged deadwood 0.00 0.22
Water hyacinth 0.00 0.99
Other floating plants 0.00 0.88
Emergent rocks 0.00 0.08
Whole-lake variables
Secchi transparency (cm) 2 5.0 130.0
Conductance ([micro]S/cm) 3 5.0 420.0
Oxygen concentration (mg/L) 3.0 12.3
pH 6 6
Surface area (ha) 2.9 250.8
Fetch (km) 0.16 3.84
Breadth (km) 0.10 1.69
Perimeter (km) 1.2 13.3
Shoreline development 1.13 2.49
Maximum depth (m) 0.45 4.60
Supra-lake variables
Distance from river channel (km) 0.2 2.9
Distance from floodplain edge (km) 0.1 2.8
Non-forested perimeter 0.00 0.80
Because water transparency, conductance, and depth can affect electrofishing effectiveness (Reynolds 1983, Zalewski and Cowx 1990), it is possible that assemblage-level patterns related to these environmental variables result from sampling artifacts, but statistical evidence suggests otherwise. Fishing success (total numerical density) was not significantly related (by Spearman rank correlation) to Secchi depth, conductance, or lake depth (Rodriguez and Lewis, unpublished data). Furthermore, our results are inconsistent with the expected direction of possible sampling biases related to visibility (Zalewski and Cowx 1990).
Fish larger than 400 mm standard length (SL) were usually returned to the water after identification and measurement. Other fish were preserved within 6 h in buffered formaldehyde (6%) and transported to the laboratory, where they were identified, measured to the nearest millimeter, and weighed to the nearest 0.1 g. Species identifications were confirmed by systematists at the Instituto de Zoologia Tropical, Universidad Central de Venezuela (UCV) in Caracas, and at the Academy of Natural Sciences of Philadelphia. Voucher specimens of all species were deposited at the Museo de Biologia UCV (MBUCV; Table 2), at the Museum of the Academy of Natural Sciences of Philadelphia (ANSP), and at the Museo de la Universidad Experimental de Los Llanos Ezequiel Zamora (MCNG), in Guanare, Venezuela.
Environmental measurements
Because factors that influence fish distribution and abundance may operate over a broad range of spatial scales, we included in the analyses sets of environmental variables differing in spatial coverage (Table 3). Each variable was assigned to one of three categories: intra-lake, whole-lake, and supra-lake. Intralake variables reflect small-scale spatial heterogeneity and measure structural attributes of microhabitats. Whole-lake variables reflect environmental features that are broadly distributed across a lake, and correspond to the notion of habitat. Supra-lake variables are physically external to the lake and have little or no direct effect on fish during the dry season, but may affect the distribution and abundance of fish at the onset of the dry season, when lakes first become isolated.
Intra-lake variables were measured along the shoreline sampling intervals (length [approximately equals] 900 m) defined by 20-min electrofishing runs. Nearshore depth was measured with a staff gauge six times per shoreline interval. Shoreline types were recorded and transformed to a fraction. Shoreline categories include: (1) submerged live terrestrial vegetation, (2) submerged wood, (3) floating grasses, (4) water hyacinth, (5) other floating plants, (6) rocks, (7) overhanging vegetation, and (8) open shoreline (usually open, gently sloping, shallow beaches).
Whole-lake variables include physico-chemical variables, which were measured at the deepest point in the lake, as follows: depth, transparency (by depth at which Secchi disk is visible), conductance (YSI model 33 meter), pH (paper strips), and dissolved oxygen (YSI model 57 meter, 0.15 m below the water surface). Also, area, fetch, width, and perimeter were measured on photographically enlarged 1:100000 maps from the Venezuelan Direccion de Cartografia Nacional (for low-water conditions). Shoreline development, a standardized measure of the relation between lake perimeter (P) and surface area (A), was calculated as P/[(4 [Pi] A).sup.0.5].
Supra-lake variables were measured on maps, and were verified in the field as necessary. Two distances describing the position of each lake on the floodplain were measured: lake centroid to main river, and lake centroid to the edge of the flooded area. The fractions of shoreline length with and without forest cover ([is greater than or equal to] 100 m away from the shoreline) were also measured.
Quantitative analyses
Species represented by [is less than] 5 individuals (53 of 169 species) were not included in analyses of species abundance. The low abundance of these species would have given them negligible weighting in the analyses even if they had been included. Species of small ([is less than] 40 mm SL; mean SL = 24 mm) characids of the subfamily Tetragonopterinae (tetras; probably fewer than eight species, mostly in the genera Astyanax, Moenkhausia, and Hyphessobrycon) that could not be identified were also excluded from analyses of relative abundance, but seven tetra species (five of [is greater than]40 mm SL; two of [is less than] 40 mm SL) that could be reliably identified were retained. As a group, the unidentified small tetras accounted for 29% by numbers but only 0.39% by mass of all fish captured in this study; their median percent numerical abundance (across lakes) was 5.6% (interquartile interval 0-36.2%). Aggregation of these species as if they were a single species would have distorted the analyses. The response of these fishes to environmental variation was examined separately.
Raw catch data were converted to catch per unit effort (CPUE = number of fishes caught divided by shocker time). The CPUE data for individual species were transformed as ln(X + 1) because an analysis of the power relationship between the means and variances of species across sites indicated that this transformation was appropriate for stabilizing variances (Rodriguez and Lewis 1994). When necessary, environmental variables were transformed as ln(X) or, for proportions, as arcsine ([p.sup.0.5]) to make the data conform better to statistical assumptions.
Statistical associations between assemblage structure and environmental variables were quantified by use of canonical correspondence analysis (CCA), a nonlinear eigenvector ordination technique especially designed for direct analysis of the relationships between multivariate ecological data sets (ter Braak 1986; program CANOCO version 2. 1, ter Braak 1988). Separate analyses were conducted for each combination of season (early and late dry season) and year (1987 and 1988), and for all samples combined. Surface area was not included among the environmental variables in the late dry season because accurate estimates of it were not available for these dates. Significance tests for models relating assemblage structure to environmental variables were based on Monte Carlo permutation tests ([10.sup.3] permutations) for the sum of all eigenvalues. Variables entering final models were selected by a stepwise procedure analogous to backward elimination in multiple regression analysis (Draper and Smith 1981). The significance of relationships between ordination axes and individual environmental variables was evaluated by t tests for the inter-set correlations and the canonical coefficients (ter Braak 1988, 1990). The weighted average site scores (ter Braak 1988) were used in the ordination plot of lake samples.
A two-part discriminant function analysis (Wilkinson 1988: SYSTAT program version 5.03) was used to determine the degree to which categories of lakes defined by environmental data corresponded to two aspects of assemblage structure: the numerical densities of major taxa, and the numerical densities of piscivorous species. In the first part of the analysis, CPUE for each lake was summed across species for each of six major taxonomic groups: siluriforms (catfishes), characiforms (e.g., silver dollars, hatchetfishes, curimatas), cichlids (e.g., peacock bass, earth eaters), clupeomorphs (anchovies and herrings), gymnotiforms (knifefishes), and other taxa (Table 2). Log-transformed CPUE totals for the six groups were then used as predictor variables for a discriminant function. In the second part of the analysis, log-transformed CPUE values for each of the 20 most abundant piscivorous species were used as predictors.
Scatterplots were used to depict changes in relative abundance, transformed as arcsine ([p.sup.0.5]) to reduce the dependence of the variance on the mean, along an environmental gradient (Wilkinson 1988: SYSTAT program version 5.03). Jittering, i.e., adding a small amount of random noise to the data before graphing (Cleveland 1985), was used to improve visibility when many data points overlapped. Robust locally weighted regression (Cleveland 1985: LOWESS) was used to summarize patterns of dependence between the y variable and the x variable in the scatterplots.
We used Mantel tests (Mantel 1967) to evaluate whether similarities in assemblage structure among lakes were a simple function of geographical distance. Mantel's test shows whether two square matrices show a statistically significant association. In the present study, one of the matrices represented a set of geographical distances among lakes and the other matrix contained a measure of assemblage dissimilarity. Maps were used to obtain geographical distances, measured as the shortest pairwise distances "as the fish swims" between each lake and all other lakes. Two transformations of species numerical densities, standardization by site norm (Noy-Meir et al. 1975) and ln(X + 1), were used in the Mantel analyses. Assemblage dissimilarity was quantified as pairwise Euclidean distances between lakes positioned in a multidimensional space in which each axis represented the transformed abundance of a species. For each set of dissimilarities, Mantel tests were performed for each combination of year and season, and within each combination tests were done at three levels differing in spatial extent: all lakes (three geographic regions), only lakes from Ciudad Bolivar and Caicara (two regions), and only lakes from Ciudad Bolivar (the region with the largest number of lakes). Significance levels were determined by reference to a null distribution derived from a random subset of permutations of one of the original matrices (Dietz 1983). To ensure the stability of the estimated probabilities (Jackson and Somers 1989), a minimum of 2500 permutations were used to generate the random null distribution; when probability estimates approached the 5% significance level, 20 000 permutations were performed. For most simulations, significance values expressed as fractions were stable to the second decimal digit after 1500-2000 permutations.
RESULTS
Relationship between assemblage structure and environmental variables
There were no significant associations between assemblage structure and supra-lake or within-lake variables in the canonical correspondence analyses. Among the whole-lake variables, only Secchi depth, conductance, depth, and area were significantly related to assemblage structure in the early dry season (Table 4). Secchi depth, conductance, and depth also were significantly related to assemblage structure in the late dry season (lake area was unavailable for late dry season analyses). These results were similar in the two years of the study (Table 4). The strongest correlations among environmental variables were between area and other measures of lake size (fetch, breadth, and perimeter: Appendix). The correlations among the best predictors in the canonical correspondence analyses (transparency, conductance, depth, and area) were generally low, as were those of transparency, conductance, and depth with other environmental variables (Appendix).
[TABULAR DATA 4 NOT REPRODUCIBLE IN ASCII]
A CCA for all samples combined showed the overall relationships between species distributions, environmental variables, and the seasonal shifts in assemblage structure (Table 4, Fig. 1). Secchi transparency increased and conductance decreased along the first ordination axis, whereas the second axis reflected a gradient in lake depth (Table 4, Fig. 1). For most species, Axis-1 scores predicted well whether the species had higher occurrence and mean density in lakes with transparency above average (clear lakes, Secchi depth [is greater than]20 cm) or below average (turbid lakes, Secchi depth [is less than or equal to] 20 cm) (Fig. 1, Table 2). The majority of species fell within an elongated cluster (running from bottom left to top right in Fig. 1) that was closely aligned with the transparency gradient. Within this cluster, species distributions were not strongly related to conductance (Fig. 1). The conductance gradient primarily separated the main species cluster from a second, smaller cluster (lower right in Fig. 1) composed mainly of species that were common in the Caura region but rare or absent in the two other regions. For several major taxa, the midpoint of the transparency gradient (Secchi depth = 20 cm) was associated with conspicuous contrasts in species distributions (Fig. 1). All nine knifefishes species had their center of distribution, corresponding to modal or peak abundance, in turbid lakes (Fig. 1, left of origin). Catfishes also were most abundant in turbid lakes: only three of 25 catfish species were positioned above the gradient midpoint (Fig. 1, right of origin; Table 2). In contrast, cichlids (six species) were most abundant in clear lakes; all cichlid species were positioned above the gradient midpoint (Fig. 1, Table 2). The distributions of species modes for characiforms (36 of 64 species above the transparency midpoint), clupeomorphs (four of eight species), and other taxa (one of four species) were less clearly patterned relative to transparency or to the other environmental variables in the plot (Fig. 1, Table 2).
[Figure 1 ILLUSTRATION OMITTED]
The relative abundance of major taxa (CPUE summed across species for each major taxon, divided by total CPUE) was predictably related to the transparency gradient; the relative abundances of characiforms, cichlids, and clupeomorphs were highest in clear lakes, whereas catfishes, knifefishes, and other taxa (mostly the sciaenid Plagioscion squamosissimus, a piscivorous croaker) were most abundant in turbid lakes (Fig. 2, Table 5). Despite marked differences in relative abundance of major taxa between clear lakes and turbid lakes, the total numerical densities were comparable in the two lake types (Table 5). As did other characiforms, the small tetras had higher numerical densities in clear lakes than in turbid lakes (Table 5).
[Figure 2 ILLUSTRATION OMITTED]
TABLE 5. Fish abundances in clear lakes (Secchi depth > 20 cm) and turbid lakes (Secchi depth [is greater than or equal to] 20 cm) of the Orinoco floodplain. Densities are given as In (CPUE+ 1), where CPUE = catch per unit effort; n = number of lakes. Standard errors are given in parentheses.
1987
Clear Turbid
Taxon (n = 23) (n = 17)
Relative abundance
(% of total CPUE)
Characiformes 60.0 36.6
(5.4) (4.7)
Siluriformes 9.3 29.5
(2.4) (5.2)
Gymnotiformes 0.6 8.4
(0.4) (3.0)
Cichlidae 7.7 3.3
(2.0) (1.1)
Clupeiformes 21.6 18.3
(4.3) (6.5)
Others 0.8 3.9
(0.5) (1.7)
Total numerical 4.1 4.1
density (0.2) (0.2)
Density of small 2.23 1.11
tetras (0.48) (0.32)
Secchi transparency (cm) 50.0 13.5
(5.8) (1.3)
1988
Clear Turbid
Taxon (n = 17) (n = 22)
Relative abundance
(% of total CPUE)
Characiformes 66.1 40.4
(5.6) (5.0)
Siluriformes 3.9 29.5
(0.9) (4.6)
Gymnotiformes 0.0 6.6
(0.0) (2.1)
Cichlidae 8.2 4.5
(1.4) (1.1)
Clupeiformes 20.8 16.9
(6.5) (3.7)
Others 1.0 2.1
(0.7) (0.8)
Total numerical 4.4 4.3
density (0.2) (0.3)
Density of small 3.09 2.08
tetras (0.52) (0.46)
Secchi transparency (cm) 44.1 11.4
(6.4) (1.2)
Members of the piscivore guild could be separated into two associations on the basis of their position along Axis 1. The first association included eight species with positive scores on Axis 1; these were most abundant in clear lakes with low conductance. The numerically dominant species in this group were the peacock bass, Cichla orinocensis, and the pike-like characiforms Acestrorhynchus microlepis, A. nasutus, and Boulengerella lucia (Fig. 3, Table 2). The second group included 12 species with negative scores on Axis 1; these were most abundant in turbid lakes with high conductance. The numerically dominant species in this group were the cynodontid characiforms Hydrolycus scomberoides, Rhaphiodon vulpinus, and R. gibbus, the croaker P. squamosissimus, and the catfishes Sorubim lima and S. latirostris (Fig. 3, Table 2). The distributions of mean density and relative abundance of piscivorous species in relation to the transparency gradient were in agreement with their assignment to the clear or turbid associations on the basis of CCA scores (Fig. 3, Table 2).
[Figure 3 ILLUSTRATION OMITTED]
The lake plot (ordination based on all samples combined) illustrated pronounced seasonal shifts in assemblage structure and environmental variables (Fig. 4). For most lakes, the lines connecting samples from the early and late dry season were approximately parallel and were most strongly aligned with the arrow representing water transparency, indicating a common direction of seasonal change in assemblage structure and an association between such change and transparency, both in 1987 and 1988 (Fig. 4). Seasonal change in assemblage structure during the dry season involved primarily increases in the relative abundance of catfishes, knifefishes, and other major taxa, and reduction in the relative abundance of characiforms and clupeomorphs (Fig. 2 [cf. Figs. 1 and 4]; Rodriguez and Lewis 1994). Because during the dry season immigration and emigration cannot occur and reproduction is negligible, the numerical abundances of species can only decrease. Thus seasonal change in relative abundances is due to differential mortality (Rodriguez and Lewis 1994). Year-to-year variation in the lake ordination scores was small; the Pearson correlation between lake scores in 1987 and 1988 (n = 39 lakes) was r = 0.95 for Axis 1 and r = 0.79 for Axis 2.
[Figure 4 ILLUSTRATION OMITTED]
The lake ordination revealed mostly continuous variation in assemblage structure within the Caicara and Ciudad Bolivar regions but marked discontinuity between the Caura region, which had the highest scores on Axis 1, and the two other regions (Fig. 4). The position of samples relative to environmental arrows indicates that differences in assemblage structure between the Caura region and the two other regions were associated primarily with conductance and to a lesser extent with transparency (Fig. 4).
The frequency distribution of ln(transparency) is approximately uniform and the relative abundances of major taxa and of most individual species vary continuously along the transparency gradient (Figs. 2, 3). Nevertheless, the discontinuity in the distribution of species modes near the mean Secchi depth of 20 cm (Fig. 1) suggests that this is a convenient break-point for lake classification. The discriminant function analysis showed that lakes could be accurately classified into two discrete categories (clear, Secchi depth [is greater than]20 cm; or turbid, Secchi depth [is less than or equal to] 20 cm) on the basis of either the numerical density of six major taxa, or the numerical density of piscivorous species, as the two analyses yielded virtually identical results (Fig. 5). Conversely, knowing whether a lake is clear or turbid was useful in predicting which major taxa and which piscivorous species predominated in the lake. Among the major taxa, the correlation between the biotic predictors and the discriminant scores (not shown) was negative for the characiforms, the cichlids, and the clupeomorphs, and positive for the catfishes, the knifefishes, and the other taxa. In the analysis based on piscivore densities, the correlation with the discriminant scores (not shown) was negative for all species that were most abundant in clear lakes (eight species, Table 2) and positive for all species that were most abundant in turbid lakes (12 species, Table 2). In both analyses the dispersion of scores was similar across categories (Fig. 5), indicating that the critical assumption of equal variance-covariance matrices was not grossly violated (Williams 1983). Alternative discriminant analyses based on either three or four categories, corresponding to equally spaced intervals along the ln(transparency) gradient, had lower classification accuracy than the simpler analysis with two categories.
[Figure 5 ILLUSTRATION OMITTED]
Variation of assemblage structure with geographical distance between lakes
The Mantel tests based on log-transformed abundances did not reveal significant associations between geographical distance-and assemblage structure for any combination of year and season (Table 6). For the site-standardized data, the tests involving all three regions yielded significant results for the early dry season in 1987 (P = 0.046) and 1988 (P = 0.048) (Table 6). This apparent effect of geographical distance on assemblage structure was caused by differences between the Caura region and the two other regions rather than by differences between the Ciudad Bolivar and Caicara regions or differences within regions. The contrast in species composition between the Caura region and the other two regions was large relative to variation within regions, as were the geographical distances between the Caura region and the two other regions. Because Mantel's test is essentially a regression procedure (Smouse et al. 1986), large geographical distances (e.g., between regions) can influence the outcome of the test more than small distances (e.g., within regions). The average geographical distances between regions, calculated as distances between region centroids, were 194 km for Caura-Caicara, 234 km for Caura-Ciudad Bolivar, and 367 km for Caicara-Ciudad Bolivar. When the Caura region was excluded from the analysis (parts b and b' of Table 6), no effect of geographical distance on assemblage structure was detected, despite the fact that this analysis retained the two most widely separated regions. When both the Caura region and the Caicara region were excluded (parts c and c' of Table 6), no effects of geographical distance were detected within the Ciudad Bolivar region. As regions were dropped from the analysis, there was a trend toward larger significance values (Table 6). In general, geographical distance between lakes was per se a poor predictor of similarity in assemblage structure. The Mantel tests also showed that there is no strong spatial autocorrelation in assemblage structure among lakes, i.e., sites are spatially independent, and thus fulfill an important requirement of the statistical tests of relationships between assemblage structure and environmental variables (Hinch et al. 1994).
TABLE 6. Significance levels of Mantel tests evaluating the relationship between assemblage structure and geographical distance. The number of lakes is given in parentheses.
1987
1988
Early Late
Regions included dry season dry season
Tests based on log-transformed data
a) Ciudad Bolivar, 0.11 0.37
Caicara, and Caura (20) (20)
b) Ciudad Bolivar and 0.17 0.46
Caicara (16) (16)
c) Ciudad Bolivar 0.05 0.81
(12) (12)
Tests based on site-standardized data
a) Ciudad Bolivar, 0.05 0.06
Caicara, and Caura (20) (20)
b) Ciudad Bolivar and 0.11 0.20
Caicara (16) (16)
c) Ciudad Bolivar 0.67 0.28
(12) (12)
Regions included Early Late
dry season dry season
Tests based on log-transformed data
a) Ciudad Bolivar, 0.12 0.10
Caicara, and Caura (20) (19)
b) Ciudad Bolivar and 0.19 0.20
Caicara (16) (16)
c) Ciudad Bolivar 0.34 0.75
(12) (12)
Tests based on site-standardized data
a) Ciudad Bolivar, 0.05 0.38
Caicara, and Caura (20) (19)
b) Ciudad Bolivar and 0.11 0.65
Caicara (16) (16)
c) Ciudad Bolivar 0.86 0.66
(12) (12)
DISCUSSION
The results revealed an unexpectedly tight relationship between fish assemblage structure and four abiotic environmental features of Orinoco floodplain lakes: transparency, conductance, depth, and area (Table 4). Most lakes showed similar trajectories of seasonal assemblage change during isolation. This change involved reduction in the abundance of visually oriented fish relative to others concomitant with decline in transparency (Figs. 1 and 4). The classification of lakes into clear and turbid types based on major taxa was nearly identical to that based on piscivores (Fig. 5), suggesting a close association between these two aspects of assemblage structure and water transparency.
Microhabitat characteristics did not appear to influence assemblage structure during isolation, even though numerous taxa were associated with specific microhabitats (Lowe-McConnell 1975, Reid 1986, Rodriguez et al. 1990). The absence of spatial effects, as shown by the Mantel tests, and the low predictive power of the supra-lake environmental variables suggest that floodplain features surrounding the lakes, which could have affected fish movement and reshuffling over flooded terrain at high water and thus their assortment into lakes at drawdown, contributed little to assemblage structure during isolation. These results point to local processes operating at the habitat scale as major determinants of assemblage structure during the dry season.
Assemblage structure in Caicara and Ciudad Bolivar lakes varied more or less continuously along an environmental gradient defined primarily by water transparency, but a marked discontinuity associated with conductance separated Caura lakes from those in the two other regions (Fig. 4). This discontinuity was due not only to differences in relative abundances of shared species, but also to regional differences in species pools. Only 23% of all species in the Caura and Caicara regions were found in both regions, and 21% of all species in the Caura and Ciudad Bolivar regions were found in both regions (Rodriguez and Lewis 1990). In contrast, 57% of all species in the Caicara and Ciudad Bolivar regions were common to both regions (Rodriguez and Lewis 1990). The discontinuity in fish assemblage structure between the Caura region and the two others is amenable to a biogeographical interpretation based on vicariance: former connections between the Caura and Upper Orinoco drainages may have affected large portions of the Caura ichthyofauna (Chernoff et al. 1991) independently of the Caicara and Ciudad Bolivar regions. Because conductance had limited value in explaining seasonal, within-region, or between-region (Caicara vs. Ciudad Bolivar) variation in fish assemblage structure (Fig. 4), its significance in this study probably resulted incidentally from the association of water type with biogeographical zonation.
[Figure 4 ILLUSTRATION OMITTED]
A working model offish assemblage variation in Orinoco floodplain lakes
Water transparency, by setting limits to underwater visibility, largely determines which major taxa and piscivore types predominate in a given lake during the dry season. Among the major taxa, catfishes and knifefishes, which are primarily nocturnal and have sensory adaptations to low light (Bennett 1971, Hara 1971, Lowe-McConnell 1975, Fink and Fink 1979), are dominant in turbid lakes, whereas characiforms, cichlids, and clupeomorphs, which usually are diurnal and rely on vision (Lowe-McConnell 1975, Fink and Fink 1979, Goulding 1980), predominate in clear lakes. The tactile and chemical sensors in catfishes and the electric sensors in knifefishes allow these fish to forage efficiently in turbid waters, which provide refuge from visual predators (Miller 1979, Blaber and Blaber 1980, Gregory 1993). In contrast, the visually oriented characiforms, cichlids, and clupeomorphs should be more vulnerable to predation when reaction distances are low (Vinyard and O'Brien 1976, Confer et al. 1978, Bruton 1985, Fraser et al. 1987); as light intensity decreases the visual advantage switches from the prey to the predator (Cerri 1983, Howick and O'Brien 1983, Pitcher and Turner 1986). For catfishes and knifefishes, the protective role of turbidity may be particularly critical during the dry season in dish lakes with a low depth/ area ratio, where structural cover declines progressively as the shoreline recedes and fish become concentrated in a relatively featureless basin. This interpretation is consistent with the patterns of the CCA ordinations (Table 4, early dry season), for which Axis 2 reflects primarily a contrast between the assemblages of turbid dish lakes and those of clear lakes of greater relative depth. The selection procedure used in the ordinations includes individual environmental variables on the basis of their significance once the effect of other environmental variables has been partialled out (ter Braak 1988). Therefore, although transparency is correlated with lake morphometry (Hamilton and Lewis 1990), depth and area probably derive their significance in the CCA analyses from their effect on availability of cover rather than from indirect effects mediated by transparency.
Species distributions that cannot be explained on the basis of predominant sensory mode often could be interpreted in terms of other adaptations. Among the nonpiscivorous taxa, many of the visually oriented species that are abundant in turbid lakes (Table 2) were found primarily near the surface where illumination is strongest (e.g., the tetras Moenkhausia intermedia and Ctenobrycon hauxwellianus). Several of these species are morphologically specialized for feeding very near the water surface (Goulding 1980: the characiforms Triportheus angulatus, Triportheus rotundatus, Triportheus culter, and Thoracocharax stellatus, and the clupeids Rhinosardinia amazonica and Pellona flavipinnis). Others, such as the characid scale-eaters Roeboides myersi and Roeboides cf. affinis, probably have nonvisual sensory adaptations to low light; Roeboides species feed at dusk or night and in water with extremely low visibility (Sazima 1983). This may be accomplished through the use of sensory pores on the head (pit lines), which would allow nonvisual detection of prey (Sazima 1983). In fact, scale-eaters with special sensory capabilities may be more successful in very turbid waters, where they can approach visually oriented prey very closely before being detected (Sazima 1977, 1983).
Species- specific differences in mortality by piscivory underlie changes in relative abundance of species during the dry season (Rodriguez and Lewis 1994). Total mortality over the dry season, averaged across the Orinoco lakes, was [is greater than] 75% (Rodriguez and Lewis 1994). This rate is not uncommonly high for tropical floodplains, where yearly mortalities [is greater than] 90% have been found for 1-yr-old fish in several species (Welcomme 1985). In the Amazon floodplain, a very high proportion of the production by prey species is consumed by piscivorous fish (Bayley 1982). In the Orinoco lakes, fish with symptoms of disease or emaciation were very rare and terrestrial predators appeared to be of little significance; mass mortalities were never observed (Rodriguez and Lewis 1994). Although direct physiological stress can sometimes influence fish assemblages (winter anoxia and low pH in temperate lakes [Tonn and Magnuson 1982, Rahel 1984, 1986]; deoxygenation, desiccation, and hydrogen sulfide poisoning in tropical floodplain lakes [Welcomme 19851), such factors seem unrelated to assemblage variation in the Orinoco lakes. No signs of lake desiccation or high concentrations of hydrogen sulfide were found during the dry season (M. A. Rodriguez, personal observation). Percent oxygen saturation near the lake bottom fell below 25% in only four of the 79 lake samples (Rodriguez and Lewis 1994), and oxygen concentration had no significant effect in the canonical correspondence analyses (Table 4). Furthermore, there appears to be no evidence that characiforms and cichlids systematically differ from catfishes and knifefishes as regards tolerance to hypoxia or high temperature (Rodriguez and Lewis 1994). Assemblage changes could not be explained in relation to pH, which showed no detectable variation in this study (Table 3).
The foraging performance of predators depends on their physiology, morphology, and behavior under prevailing environmental conditions (Ryder 1977, Hobson 1979, Miller 1979, Hobson et al. 1981). Piscivores of the Cichla-Acestrorhynchus-Boulengerella association, which characterizes clear lakes, are visual mid-water predators that lunge forward to capture prey (Zaret 1979, 1980, Menezes et al. 1981, Nico and Taphorn 1985). In well-lighted temperate lakes, dominant visual piscivores such as pike and muskellunge (Esox spp.) and bass (Micropterus spp.), which rely for their success on high acceleration and the ability to strike from cover (Guthrie 1986), are remarkably similar in overall morphology to the clearwater piscivores of the Orinoco lakes. Boulengerella species are solitary predators that strike from hiding places in submerged vegetation, whereas Cichla and Acestrorhynchus may hunt in groups. Some field evidence shows that the effect of Cichla ocellaris on prey populations depends on water transparency (Zaret 1979).
Piscivores in the cynodontid-catfish-Plagioscion association, which were found primarily in turbid lakes, differ in foraging mode. The three cynodontids, Hydrolycus scomberoides, Rhaphiodon vulpinus, and R. gibbus, are surface-oriented predators that have remarkably similar morphology. The narrow head allows stereoscopic vision anteriorly, ventrally, and dorso-anteriorly (Howes 1976). In very turbid waters, these fish likely detect prey from below (Ali 1959, Hobson 1979). The enlarged pectoral fins of these fish probably are used for rapid upward acceleration, followed by prey capture either by impalement on large inferior canines in the upturned mouth, or by suction caused by expansion of the buccal cavity (Howes 1976, Goulding 1980). In turbid lakes, the high upwelling illumination decreases the contrast between predator and background, making predators such as these less visible (Muntz 1982).
The piscivorous catfishes, such as Pseudoplatystoma, are predominantly benthic and probably rely mostly on tactile and chemical stimuli to detect prey (Reid 1983). The croaker Plagioscion squamosissimus is a silvery fish with a large mouth and a flattened ventrum. Croakers and drums (Sciaenidae) often live in turbid waters and have elaborate sound producing and receiving systems and a well-developed lateral line (Moyle and Cech 1988). Nico and Taphorn (1984) have shown that P. squamosissimus feeds near or on the bottom during the night, and Worthmann and de Oliveira (1983) have shown that this species feeds well in turbid lakes of the Amazon.
The model of fish assemblage structure that emerges from the associations of species groups with environmental variables involves causal relationships connecting piscivory to transparency and transparency to lake morphometry. Piscivores modify assemblage structure during the dry season by culling the most vulnerable prey species. Vulnerability in turn is strongly related to transparency. Transparency at low water is predictably related to the depth and size of lakes: resuspension of sediment leading to turbidity occurs at specific thresholds of depth and fetch (Hamilton and Lewis 1990). In addition to its role as a determinant of transparency, lake morphometry affects vulnerability through the relationship between lake relative depth and availability of structural cover. Furthermore, prey species may be able to use morphometric features of lakes as indicators of predation intensity, and predators may select lakes that are most compatible with their mode of predation. This model emphasizes the interaction between lake transparency and the properties of individual species, in particular their sensory and foraging capabilities. Assemblage structure is strongly deterministic because lake morphometry channelizes the outcome of interactions between predators and prey in the same way from year to year. Thus, lakes with similar environmental features have similar fish assemblages, and such similarities are largely independent of geographical distance between lakes.
Generality of the model
The piscivory-transparency-morphometry (PTM) model may be applicable to other tropical floodplains. Turbid whitewaters are usually richer in nutrients and less acidic than more transparent clearwaters and blackwaters. Accordingly, Marlier (1967) and Goulding (1980) have stressed the effects of water chemistry and productivity when interpreting the marked faunal differences between these water types (Goulding and Ferreira 1984, Lowe-McConnell 1987). The PTM model may provide a more accurate description of influences on fish assemblages, however.
The PTM model is rich in testable predictions and suggests lines for further research at several levels. The environmental variables that proved to be significant in this study can be measured easily for other floodplains. Behavior and feeding performance of fishes can be assessed in experiments that involve manipulation of visibility. Physiological and behavioral studies can determine whether specific adaptations match the assumptions of the model. For example, the retinas of characiforms generally absorb at longer wavelength than those of cichlids and presumably confer increased sensitivity to characiforms in turbid waters (Muntz 1982), which may explain why characiforms are less restricted to transparent lakes than cichlids (Fig. 1).
The PTM model provides context for questions about fish adaptation and evolution. If trophic interactions are a major selective force whose outcome depends on the visual environment (Hobson et al. 1981), how does this phenomenon affect coloration patterns, microhabitat preferences, and selection of spawning sites? Does the timing of ontogenetic shifts differ among lakes with contrasting transparency? How do small zooplanktivores avoid predators while foraging in open waters and how do their adaptations depend on the optical environment?
To date, most extensive multi-lake surveys relating fish assemblage structure to lake features have been conducted in temperate zones. These studies have pointed to a biotic factor, piscivory, various abiotic factors such as area, depth, productivity, pH, alkalinity, winter hypoxia, and conductance, and biogeographic factors, including lake isolation or connectivity, as major determinants of fish assemblage structure (Johnson et al. 1977, Tonn and Magnuson 1982, Rahel 1984, 1986, Marshall and Ryan 1987, Jackson and Harvey 1989, Robinson and Tonn 1989, Tonn et al. 1990, Hinch et al. 1991, Jackson et al. 1992, Hinch et al. 1994). Although numerous studies have documented a marked influence of the underwater optical environment on individual foraging performance (e.g., Ali 1959, Ryder 1977, Fraser et al. 1987, McMahon and Holanov 1995), competition between species pairs (Bergman 1991) ' and spatial distributions of species (estuaries: Blaber and Blaber 1980, coral reefs: Roberts 1991), reports of a significant correlation between lake transparency and the structure of whole fish assemblages are rare (e.g., Johnson et al. 1977, Marshall and Ryan 1987). Marshall and Ryan (1987) attribute this correlation to a link between transparency and lake productivity without referring to a possible direct effect of transparency. In one of the few studies emphasizing the synecological effects of subsurface illumination in lakes, Ryder and Kerr (1989) discuss how light levels influence the geographic distributions of fish species, such as the congeners walleye (Stizostedion vitreum) and sauger (S. canadense), and suggest that competitive relationships between some species pairs may be determined by the light regime. The present study suggests that greater emphasis on the transparency-piscivory linkage may prove fruitful in explaining the distribution and abundance of fishes.
CONCLUSIONS
In agreement with the findings of Robinson and Tonn (1989), the fish assemblages of Orinoco floodplain lakes appear to be structured by a small number of environmental variables and one biotic variable, predation. Although water transparency has not been strongly linked to assemblage structure in earlier studies, transparency alone is a remarkably reliable predictor of which species, major taxa, and piscivore associations predominate in Orinoco lakes, and may explain correlations between environmental variables and fish assemblage structure in other studies. Fish with sensory adaptations to low light are dominant in turbid lakes, whereas fish relying on vision predominate in clear lakes, suggesting that the effect of transparency on assemblage structure is mediated by the constraints that the underwater optical environment places on foraging efficacy. The outcome of predator-prey interactions in floodplain lakes, and possibly other kinds of lakes as well, is dictated by transparency, which is predictable from morphometry. Thus fish assemblage structure is deterministic through relatively simple mechanisms on the Orinoco floodplain; a search for similar mechanisms elsewhere may prove fruitful.
ACKNOWLEDGMENTS
S. Richardson provided valuable assistance in the field under sometimes trying circumstances. G. Cagnino's skillful assistance greatly expedited the lab work. R. Royero, F. Provenzano, J. Lundberg, W. Saul, S. Schaefer, and L. Rapp Py-Daniel kindly helped with fish systematics. P. Bayley, J. Lundberg, and M. Power provided useful feedback at various stages of this project. We are grateful to F. J. Rahel and an anonymous reviewer for their helpful comments and suggestions. L. Persson offered useful pointers to the percid literature. MAR thanks his family for providing support throughout. We thank the Venezuelan Ministerio del Ambiente y de los Recursos Naturales Renovables for logistical assistance. Financial support came from the Natural Sciences and Engineering Research Council of Canada, the National Geographic Society, the National Science Foundation, and the University of Colorado.
LITERATURE CITED
Ali, M. A. 1959. The ocular structure, retinomotor and photobehavioral responses of juvenile Pacific salmon. Canadian Journal of Zoology 37:965-996.
Bayley, P. B. 1982. Central Amazon fish populations: biomass, production and some dynamic characteristics. Dissertation. Dalhousie University, Nova Scotia, Canada.
Bennett, M. V. L. 1971. Electric organs. Pages 347-491 in W. S. Hoar and D. J. Randall, editors. Fish physiology. Volume 5. Academic Press, New York, New York, USA.
Bergman, E. 1991. Changes in abundance of two percids, Perca fluviatilis and Gymnocephalus cernuus, along a productivity gradient: relations to feeding strategies and competitive abilities. Canadian Journal of Fisheries and Aquatic Sciences 48:536-545.
Blaber, S. J. M., and T. G. Blaber. 1980. Factors affecting the distribution of juvenile estuarine and inshore fish. Journal of Fish Biology 17:143-162.
Bonetto, A., E. Cordiviola de Yuan, C. Pignalberi, and O. Oliveros. 1970a. Nuevos aportes al conocimiento de las poblaciones icticas en cuencas temporarias del valle de inundacion del Parana Medio. Acta Zoologica Lilloana 27: 135-153.
Bonetto, A. A., E. Cordiviola de Yuan, and C. Pignalberi. 1970b. Nuevos datos sobre poblaciones de peces en ambientes leniticos permanentes del Parana Medio. Physis 30: 141-154.
Bruton, M. N. 1985. The effects of suspensoids on fish. Hydrobiologia 125:221-241.
Cerri, R. D. 1983. The effect of light intensity on predator and prey behavior in cyprinid fish: factors that influence prey risk. Animal Behaviour 31:736-742.
Chernoff, B., A. Machado-Allison, and W. G. Saul. 1991. Morphology, variation and biogeography of Leporinus brunneus Myers (Pisces: Characiformes: Anostomidae). Ichthyological Exploration of Freshwaters 1:295-306.
Cleveland, W. S. 1985. The elements of graphing data. Wadsworth, Pacific Grove, California, USA.
Confer, J. L., G. L. Howick, M. H. Corzette, S. L. Kramer, S. Fitzgibbon, and R. Landesberg. 1978. Visual predation by planktivores. Oikos 31:27-37.
Dietz, E. J. 1983. Permutation tests for association between two distance matrices. Systematic Zoology 32:21-26.
Draper, N., and H. Smith. 1981. Applied regression analysis. John Wiley and Sons, New York, New York, USA.
Fink, W. L., and S. V. Fink. 1979. Central Amazonia and its fish. Comparative Biochemistry and Physiology A 62:13-29.
Fraser, D. F, D. A. DiMattia, and J. D. Duncan. 1987. Living among predators: the response of a stream minnow to the hazard of predation. Pages 121-127 in W. J. Matthews and D. C. Heins, editors. Community and evolutionary ecology of North American stream fishes. University of Oklahoma Press, Norman, Oklahoma, USA.
Goulding, M. 1980. The fishes and the forest, explorations in Amazonian natural history. University of California Press, Berkeley, California, USA.
Goulding, M., and E. F. G. Ferreira. 1984. Shrimp-eating fishes and a case of prey-switching in Amazon rivers. Revista Brasileira de Biologia 2:85-97.
Goulding, M., M. L. Carvalho, and E. G. Ferreira. 1988. Rio Negro, rich life in poor waters. SPB Academic, The Hague, The Netherlands.
Gregory, R. S. 1993. Effect of turbidity on the predator avoidance behavior of juvenile chinook salmon (Oncorhynchus tshawytscha). Canadian Journal of Fisheries and Aquatic Sciences 50:241-246.
Guthrie, D. M. 1986. Role of vision in fish behaviour. Pages 75-113 in I J. Pitcher, editor. The behavior of teleost fishes. Johns Hopkins University Press, Baltimore, Maryland, USA.
Hamilton, S. K., and W M. Lewis, Jr. 1987. Causes of seasonality in the chemistry of a lake on the Orinoco River floodplain, Venezuela. Limnology and Oceanography 32: 1277-1290.
Hamilton, S. K., and W. M. Lewis, Jr. 1990. Basin morphology in relation to chemical and ecological characteristics of lakes on the Orinoco River floodplain, Venezuela. Archiv fur Hydrobiologie 119:393-425.
Hara, T. J. 1971. Chemoreception. Pages 79-120 in W.S. Hoar and D.J. Randall, editors. Fish physiology. Volume 5. Academic Press, New York, New York, USA.
Hinch, S. C. 1991. Small- and large-scale studies in fisheries ecology: the need for cooperation among researchers. Fisheries 16:22-27.
Hinch, S. G., N. C. Collins, and H. H, Harvey. 1991. Relative abundance of littoral zone fishes: biotic interactions, abiotic factors, and postglacial colonization. Ecology 72:1314-1324.
Hinch, S. G., K. M. Somers, and N. C. Collins. 1994. Spatial autocorrelation and assessment of habitat-abundance relationships in littoral zone fish. Canadian Journal of Fisheries and Aquatic Sciences 51:701-712.
Hobson, E. S. 1979. Interactions between piscivorous fishes and their prey. Pages 231-242 in H. Clepper, editor' Predator-prey systems in fisheries management. Sport Fishing Institute, Washington, D.C., USA.
Hobson, E. S., W. N. McFarland, and J. R. Chess. 1981. Crepuscular and nocturnal activities of Californian nearshore fishes, with consideration of their scotopic visual pigments and the photic environment. Fishery Bulletin 79: 1-30.
Howes, G. J. 1976. The cranial musculature and taxonomy of characoid fishes of the tribes Cynodontini and Characini. Bulletin of the British Museum of Natural History (Zoology) 29:203-248.
Howick, G. L., and W J. O'Brien. 1983. Piscivorous feeding behavior of largemouth bass: an experimental analysis. Transactions of the American Fisheries Society 112:508-516.
Jackson, D. A., and K. M. Somers. 1989. Are probability estimates from the permutation model of Mantel's test stable? Canadian Journal of Zoology 67:766-769.
Jackson, D. A., and H. H. Harvey. 1989. Biogeographic associations in fish assemblages: local vs. regional processes. Ecology 70:1472-1484.
Jackson, D. A., K. M. Somers, and H. H. Harvey. 1992. Null models and fish communities: evidence of nonrandom patterns. American Naturalist 139:930-951.
Johnson, M. G., J. H. Leach, C. K. Minns, and C. H. Olver. 1977. Limnological characteristics of Ontario lakes in relation to associations of walleye (Stizostedion vitreum vitreum), northern pike (Esox lucius), lake trout (Salvelinus namaycush), and smallmouth bass (Micropterus dolomieui). Journal of the Fisheries Research Board of Canada 34: 1592-1601.
Junk, W J., G. M. Soares, and F. M. Carvalho. 1983. Distribution of fish species in a lake of the Amazon River floodplain near Manaus (Lago Camaleao), with special reference to extreme oxygen conditions. Amazoniana 7:397-431.
Kodric-Brown, A., and J. H. Brown. 1993. Highly structured fish communities in Australian desert springs. Ecology 74: 1847-1855.
Lagler, K. F. 1978. Capture, sampling and examination of fishes. Pages 7-47 in T. F. Bagenal, editor. Methods for the assessment of fish production in fresh waters. Third edition. Blackwell Scientific, Oxford, England.
Lowe-McConnell, R. H. 1975. Fish communities in tropical freshwaters. Longman, New York, New York, USA.
--. 1987. Ecological studies in tropical fish communities. Cambridge University Press, Cambridge, England.
Mago-Leccia, F. 1970. Estudios preliminares sobre la ecologia de los peces de los llanos de Venezuela. Acta Biologica Venezuelica 7:71-102.
Mantel, N. A. 1967. The detection of disease clustering and a generalized regression approach. Cancer Research 27: 209-220.
Marlier, G. 1967. Ecological studies on some takes of the Amazon Valley. Amazoniana 1:91-115.
Marshall, T. R., and P. A. Ryan. 1987. Abundance patterns and community attributes of fishes relative to environmental gradients. Canadian Journal of Fisheries and Aquatic Sciences 44 (Supplement 2):198-215.
McMahon, T. E, and S. H. Holanov. 1995. Foraging success of largemouth bass at different light intensities: implications for time and depth of foraging. Journal of Fish Biology 46:759-767.
Menezes, N. A., H. J. Wagner, and M. A. Ali. 1981. Retinal adaptations in fishes from a floodplain environment in the Central Amazon Basin. Revue Canadienne de Biologie 40: 111-132.
Miller, R. J. 1979. Relationships between habitat and feeding mechanisms in fishes. Pages 269-280 in H. Clepper, editor. Predator-prey systems in fisheries management. Sport Fishing Institute, Washington, D.C., USA.
Moyle, P. B., and J. J. Cech, Jr. 1988. Fishes: an introduction to ichthyology. Prentice-Hall, Englewood Cliffs, New Jersey, USA.
Muntz, W. R. A. 1982. Visual adaptations to different light environments in Amazonian fishes. Revue Canadienne de Biologie Experimentale 41:35-36.
Nico, L. G., and D. C. Taphorn. 1984. Biologia de la curvinata, Plagioscion squamosissimus en el modulo "Fernando Corrales" de la UNELLEZ, Estado Apure. Revista Universidad Experimental de los Llanos Occidentales "Ezequiel Zamora" de Ciencia y Tecnologia, Barinas, Venezuela 2:31-40.
Nico, L. G., and D. C. Taphorn. 1985. Diet of Acestrorhynchus microlepis (Pisces: Characidae) in the low llanos of Venezuela. Copeia 1985:794-796.
Novotny, D. W., and G. R. Priegel. 1974. Electrofishing boats. Improved designs and operational guidelines to increase the effectiveness of boom shockers. Technical Bulletin Number 73. Department of Natural Resources, Madison, Wisconsin, USA.
Noy-Meir, L, D. Walker, and W. I Williams. 1975. Data transformations in ecological ordination. 1. On the meaning of data standardization. Journal of Ecology 63:779-800.
Pitcher, T. J., and J. R. Turner. 1986. Danger at dawn: experimental support for the twilight hypothesis in shoaling minnows. Journal of Fish Biology 29A:59-70.
Rahel, F J. 1984. Factors structuring fish assemblages along a bog lake successional gradient. Ecology 65:1276-1289.
--. 1986. Biogeographic influences on fish species composition of northern Wisconsin lakes with applications for lake acidification studies. Canadian Journal of Fisheries and Aquatic Sciences 43:124-134.
Rahel, F J., and W. A. Hubert. 1991. Fish assemblages and habitat gradients in a Rocky Mountain-Great Plains stream: biotic zonation and additive patterns of community change. Transactions of the American Fisheries Society 120:319-332.
Reid, S. 1983. La Biologia de los bagres rayados Pseudoplatystoma fasciatum y P. tigrinum en la cuenca del Rio Apure, Venezuela. Revista Universidad Experimental de los Llanos Occidentales "Ezequiel Zamora" de Ciencia y Tecnologia, Barinas, Venezuela 1: 13-41.
--. 1986. Cryptic adaptations of small juvenile catfishes Sorubim lima (Pimeladidae) in Venezuela. Biotropica 18: 86-88.
Reynolds, J. B. 1983. Electrofishing. Pages 147-163 in L.A. Nielsen and D.L. Johnson, editors. Fisheries techniques. American Fisheries Society, Bethesda, Maryland, USA.
Roberts, C. M. 1991. Larval mortality rate and the composition of coral reef fish communities. Trends in Ecology and Evolution 6:83-87.
Robinson, C. L. K., and W. M. Tonn. 1989. Influence of environmental factors and piscivory in structuring fish assemblages of small Alberta lakes. Canadian Journal of Fisheries and Aquatic Sciences 46:81-89.
Rodriguez, M. A., and W. M. Lewis, Jr. 1990. Diversity and species composition of fish communities of Orinoco floodplain lakes. National Geographic Research 6:319-328.
Rodriguez, M. A., and W. M. Lewis, Jr. 1994. Regulation and stability in fish assemblages of neotropical floodplain lakes. Oecologia 99:166-180.
Rodriguez, M. A., S. E. Richardson, and W. M. Lewis, Jr. 1990. Nocturnal behavior and aspects of the ecology of a driftwood catfish, Entomocorus gameroi (Auchenipteridae). Biotropica 22:435-438.
Ryder, R. A. 1977. Effects of ambient light variations on behavior of yearling, subadult, and adult walleyes (Stizostedion vitreum vitreum). Journal of the Fisheries Research Board of Canada 34:1481-1491.
Ryder, R. A., and S. R. Kerr. 1989. Environmental priorities: placing habitat in hierarchic perspective. Pages 2-12 in C. D. Levings, L. B. Holtby, and M. A. Henderson, editors. Proceedings of the National Workshop on Effects of Habitat Alteration on Salmonid Stocks. Canadian Special Publication in Fisheries and Aquatic Sciences 105.
Sazima, I. 1977. Possible causes of aggressive mimicry in a neotropical scale-eating fish. Nature 270:510-512.
--. 1983. Scale-eating in characoids and other fishes. Environmental Biology of Fishes 9:87-101.
Schlosser, I. J. 1987. A conceptual framework for fish communities in small warmwater streams. Pages 17-24 in W.J. Matthews and D. C. Heins, editors. Community and evolutionary ecology of North American stream fishes. University of Oklahoma Press, Norman, Oklahoma, USA.
Smouse, P. E., J. C. Long, and R. R. Sokal. 1986. Multiple regression and correlation extensions of the Mantel test of matrix correspondence. Systematic Zoology 35: 627-632.
ter Braak, C. J. F. 1986. Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67:1167-1179.
--. 1988. CANOCO -a FORTRAN program for canonical community ordination by [partial] [detrended] [canonical] correspondence analysis and redundancy analysis (version 2.1). Groep Landbouwwiskunde, Wageningen, The Netherlands.
--. 1990. Interpreting canonical correlation analysis through biplots of structure correlations and weights. Psychometrika 55:519-531.
Tonn, W. M., and J. J. Magnuson. 1982. Patterns in the species composition and richness of fish assemblages in northern Wisconsin lakes. Ecology 63:1149-1166.
Tonn, W. M., J. J. Magnuson, M. Rask, and J. Toivonen. 1990. Intercontinental comparison of small-lake fish assemblages: the balance between local and regional processes. American Naturalist 136:345-375.
Tonn, W. M., C. A. Paszkowski, and 1. J. Holopainen. 1992. Piscivory and recruitment: mechanisms structuring prey populations in small lakes. Ecology 73:951-958.
Vinyard, G. L., and W. J. O'Brien. 1976. Effects of light and turbidity on the reactive distance of bluegill (Lepomis macrochirus). Journal of the Fisheries Research Board of Canada 33:2845-2849.
Welcomme, R. L. 1979. Fisheries ecology of flooplain rivers. Longmann, New York, New York, USA.
--. 1985. River fisheries. FAO Technical Paper 262, Rome, Italy.
Wilkinson, L. 1988. SYSTAT: the system for statistics. SYSTAT, Evanston, Illinois, USA.
Williams, B. K. 1983. Some observations on the use of discriminant analysis in ecology. Ecology 64:1283-1291.
Worthmann, H., and J. L. de Oliveira. 1983. Comparative nutritional analysis of two sciaenidian species, the pescadas, Plagioscion squamosissimus Heckel and Plagioscion monti Snares, from different water systems of the central Amazon. Animal Research and Development 21:7-34.
Zalewski, M., and I. G. Cowx. 1990. Factors affecting the efficiency of electric fishing. Pages 89-111 in I.G. Cowx and P. Lamarque, editors. Fishing with electricity. Fishing News Books, London, England.
Zaret, T. M. 1979. Predation in freshwater fish communities. Pages 135-144 in H. Clepper, editor. Predator-prey systems in fisheries management. Sport Fishing Institute, Washington, D.C., USA.
Zaret, T. M. 1980. Life history and growth relationships of Cichla ocellaris, a predatory South American cichlid. Biotropica 12:144-157.
APPENDIX
Matrix of linear correlation coefficients for transformed environmental variables.
a b c d e f g h
b -0.33
c 0.16 -0.09
d 0.16 -0.51 0.70
e -0.08 -0.29 0.61 0.91
f 0.30 -0.51 0.59 0.92 0.72
g -0.01 -0.36 0.62 0.96 0.95 0.84
h -0.52 0.35 0.08 0.24 0.50 0.06 0.50
i 0.41 -0.05 -0.03 -0.02 0.07 0.06 0.05 0.21
j 0.53 0.08 -0.08 -0.33 -0.35 -0.14 -0.33 -0.16
k -0.05 -0.49 0.21 0.56 0.52 0.56 0.53 0.05
l -0.13 -0.01 0.25 0.47 0.55 0.37 0.52 0.42
m -0.27 0.63 -0.24 -0.44 -0.33 -0.36 -0.31 0.25
n 0.39 -0.36 0.12 0.42 0.37 0.26 0.37 0.09
o 0.34 -0.37 0.08 0.08 0.00 0.04 -0.02 -0.24
P 0.07 0.10 -0.14 0.06 -0.03 0.13 -0.02 -0.31
q -0.44 0.42 0.07 -0.30 -0.17 -0.27 -0.19 0.29
r 0.10 0.21 0.21 0.01 0.11 0.00 0.09 0.28
s -0.23 0.02 -0.34 0.08 0.06 0.22 0.17 0.24
t -0.00 0.14 -0.05 -0.08 -0.10 -0.08 -0.10 -0.11
u 0.04 0.12 0.34 -0.06 -0.09 -0.01 -0.07 -0.01
i j k l n o p
b
c
d
e
f
g
h
i
j 0.70
k -0.04 -0.20
l 0.05 -0.31 0.32
m 0.03 0.21 -0.58 -0.06
n 0.20 0.17 0.15 0.19 -0.37
o 0.16 0.11 0.25 -0.03 -0.50 0.23
P 0.06 0.17 -0.34 -0.42 0.25 -0.14 -0.33
q -0.32 -0.34 -0.08 0.34 0.18 -0.57 -0.30 -0.35
r 0.32 0.19 0.26 0.27 -0.06 0.10 0.31 -0.30
s 0.13 0.10 -0.01 0.01 0.48 -0.24 -0.36 0.14
t -0.14 -0.06 -0.12 -0.54 0.17 -0.06 -0.12 0.49
u 0.09 -0.01 -0.15 -0.37 0.02 -0.06 0.19 -0.11
q r s u
b
c
d
e
f
g
h
k
l
m
n
o
P
q
r -0.12
s -0.18 -0.09
t -0.23 -0.05 -0.02
u 0.03 0.13 -0.14 0.05
Notes: a = Water transparency, b = Conductance c = Oxygen concentration; d = Surface area; e = Fetch; f = Breadth; g = Lake perimeter; h = Shoreline development; i = Maximum depth; j = Nearshore depth; k = Distance from river; l = Distance from floodplain edge; m = Non-forested perimeter; n = Submerged vegetation; o = Overhanging vegetation; p = Floating grasses; q = Open shoreline; r = Submerged deadwood; s = Water hyacinth; t = Other floating plants; u = Emergent rocks.
Shoreline types (variables n through u) were not measured in early 1987. When necessary, correlations for these variables were calculated by pairwise deletion. pH was not included in the table because it showed no detectable variation.
Manuscript received 21 July 1995; revised I March 1996; accepted 4 April 1996.
Marco A. Rodrigues(1) and William M. Lewis, Jr. Department of Evironmental, Population, and Organismic Biology, and Center for Limnology, University of Colorado, Boulder, Colorado 80309-0334 USA
(1) Present address: Departement de biologie et des sciences de la sante, Universite du Quebec A Rimouski, C.P. 3300, Rimouski, Quebec Canada G5L 3A1.
 Printer friendly
 Cite/link
 Email
 Feedback
| |
| Author: | Rodriguez, Marco A.; Lewis, William M., Jr. |
|---|---|
| Publication: | Ecological Monographs |
| Date: | Feb 1, 1997 |
| Words: | 11197 |
| Previous Article: | Patterns of predation on noncyclic lemmings. |
| Next Article: | Hatching asynchrony, brood reduction, and food limitation in a neotropical parrot. |
| Topics: | |