Spatial and Temporal Variation in Tropical Fish Trophic Networks

Observed properties of natural food webs have both important theoretical and important management implications. Four lowland aquatic food webs were investigated over the course of two years: a large swamp and a small stream in Costa Rica, and a similar swamp and stream in the Venezuelan llanos. Each local ecosystem differed from the three others with respect to environmental changes associated with seasonal rainfall. Phylogenetic composition and diversity of biotas also varied among systems. Volumetric proportional utilization coefficients from fish gut contents were used as estimates of the intensity of predator—prey interactions. An annual and two or more seasonal food webs were constructed for each local community. Aquatic communities were defined operationally using common fish species as consumers, and using the sink subweb associated with the top predator of each system. A computer calculated a variety of food—web statistics and plotted food—web diagrams containing either (a) all observed trophic links (predator—prey interactions), or (b) subsets with weak links eliminated at prescribed thresholds. Individual community food webs contained from 58 (stream, Costa Rica) to 104 (swamp, Venezuela) interactive taxonomic units and from 208 to 1243 total trophic links. Food—web parameters were very sensitive to changes in level of link threshold. Web connectance and related parameters converged near link threshold 0.04 (utilization coefficients <0.04 eliminated) in a variety of inter—web comparisons. Despite large differences in assemblage composition and attributes of the physical environment, distributions of trophic levels calculated according to a trophic continuum algorithm were very similar among study systems. Herbivores, detritivores, and their direct predators formed the largest proportions of fishes in each assemblage, followed by omnivores and secondary carnivores. Fishes that fed at more than one trophic interval were extremely common in all food webs. Analysis of covariance was used to compare structural features of different webs across a range of link thresholds. Extensive among—site variation in food—web parameters was associated with differences in species richness and environmental differences associated with rainfall patterns, physiography, and gross primary production. Seasons generally influenced food—web parameters less than did site differences. Relative importance of detritus, aquatic primary production, and terrestrial production in aquatic food webs varied seasonally in each system. Detritus, derived primarily from aquatic macrophytes, was an important pathway in both tropical swamp ecosystems. Aquatic primary productivity comprised the largest fraction of fish diets during the wet season in the Venezuelan swamp, but it formed the major component of fish diets during the dry season at all other sites. Based on comparisons using 13 webs, two—thirds of the pairings among six food—web parameters used (number of nodes, compartmentation, connectance, average number of prey per node, average number of predators per node, ratio of consumer nodes to total nodes) were positively intercorrelated. Several food—web relationships previously described as constant (e.g., connectance x species richness constancy, species scaling law, link—species scaling law) were not confirmed by my data. These earlier food—web trends are extremely sensitive to methodological biases, especially decisions regarding the degree of taxonomic lumping of species into trophic units. Although food webs have unique emergent properties and spinoff a number of potentially informative macrodescriptors, empirical studies must achieve greater precision and uniformity before analyses can be performed across different systems. Several problems and potential resolutions are discussed.