- University of Maryland
Summary
Food webs for 15 freshwater ecosystems in North America are reconstructed, based upon fish feeding habits in these ecosystems and established trophic categorizations for aquatic invertebrates. My own research in Goose Creek, Virginia, as well as literature data, provided the necessary information. One property of theoretical interest for food webs, namely omnivory, is examined, and the results are related to various topics more typically studied by freshwater fish ecologists, including predator-prey regulation, optimal foraging theory, and tropho-dynamics. The data indicate that omnivory is often important in freshwater ecosystems, such that fish may not always control the abundance of their prey. It is emphasized that knowledge of omnivory patterns can provide important insights into community structure and function in freshwater ecosystems.
Methodology
The 15 freshwater ecosystems compared in the present paper included trophic data collected by myself (Table 1; see also Table 2 and Vadas 1988), as well as literature data for 14 other watersheds (see Tables 3–16). These 15 ecosystems were chosen for food web analysis for the following reasons: (1) Several of the common fish species and/or size classes of fishes were studied. (2) Sample sizes for fishes subjected to gut analysis usually were large. (3) Food organisms were usually identified down to similar taxonomic levels (ordinal or familial level for most insects). (4) Food data were given in volumetric (or gravimetric) percentages. Certain modifications of the data, however, were necessary, as discussed below. (5) The food data were collected over space and/or time, providing some replication; e.g., I collected fishes from July to September in 1986, at four locations within a 5 km stretch of stream. If the author(s) did not pool their spatiotemporally replicated data, I did so myself, using unweighted averages. (6) Adult and juvenile fishes often were separated in the gut analyses, so that life history omnivory (see above) could be taken into account in the present analysis. Several of the studies, including my own (Vadas 1988), examined only forage fishes, and thus did not study adults of longer-lived fish species. Other studies focused upon game fishes, especially the larger size classes. Many of the studies divided fish species into two or more size classes; e.g., I divided the most abundant fish species of Goose Creek, namely the shiner, Notropis procne, into four size classes (Vadas 1988). Except in the following two cases, I used the size class designations of the author(s). First, I combined the two size classes of Hiodon tergisus and Labidesthes sicculus in Boesel's (1937) data set, to increase sample sizes and to enhance the temporal spread of the data. Second, I combined the smallest two size classes of Micropterus salmoides in the data set of Beaver and Bull Shoals Reservoirs, because the researchers did not always separate these size classes (Applegate et al. 1966, Applegate & Mullan 1967, Mullan & Applegate 1967). The above separations of fish species into size classes are justifiable, and often preferable, for food web analyses (Cohen 1978, Pimm & Rice 1987). I consider these individual, separated size classes to be trophospecies (ecological species; sensu Cohen 1978) in the present analysis.
There were four necessary modifications of the trophic data for the lotic ecosystems. First, McNeely (1986) considered sand to be a gut content item, so I recalculated the volumetric percentages of food items for McNeely's fish species to exclude sand. Second, although Schreiber & Minckley (1981) tabulated their food data in two ways, I used their percent-of-total-food-volume data in the present paper. Third, I used the percent-total-weight data of Northcote et al. (1979), although these researchers tabulated their trophic data in four other ways. Fourth, Northcote et al. (1979) did not weigh plant foods, so I had to estimate plant contributions to the diet. I did this by averaging the percent-occurrence, percent-total-number, and percent-average-number data for these foods, adjusting the percentages of animal prey to maintain 100% cumulative totals.
There were five necessary modifications of the trophic data for natural lake ecosystems. First, individual species of prey fishes in Keast's (1985, Table 1) data were tabulated separately from the invertebrate and overall fish prey (Keast 1978, Table 5), so I combined the two data sets. For example, Micropterus salmoides included fish as 82% of its overall diet (by volume), and Perca flavescens as 17% of its fish diet, by number. Hence, the overall volumetric percentage for P. flavescens prey was (82%) (17%)/100 = 14%, using the reasonable assumption that different species of fish prey were of similar average volume (size). Second, I found it necessary to lump Keast's cyprinid, as well as centrarchid species together, because of the large number of unidentified minnows and sunfishes. Third, I tabulated volumetric percentages from mean prey volumes (in cubic mm) given by Sadzikowski & Wallace (1976), after completely removing the large volumes of unidentified insects and other animals from the data set. Fourth, I split Sadzokowski & Wallace's 'general miscellaneous' food category into 50% fish eggs and 50% detritus, as the authors did not indicate the relative abundance of these foods. Finally, although Ewers (1933) and Ewers & Boesel (1935) identified several prey items down to low taxonomic levels (e.g., species level for most microcrustaceans), the presence of several unidentified prey at these same taxonomic levels necessitated the use of pooled categories for the present analysis. This was the genus level for most microcrustaceans, and other coalesced categories included fish, damselflies, hymenopterans, Tanypodinae chironomids, and other chironomids.