Running Head : Trophic Levels and Omnivory in Large Food Webs LIMITS TO TROPHIC LEVELS AND OMNIVORY IN COMPLEX FOOD WEBS : THEORY AND DATA RICHARD

While trophic levels have found broad application throughout ecology, they are also in much contention on analytical and empirical grounds. Here, we use a new generation of data and theory to examine long-standing questions about trophic-level limits and degrees of omnivory. The data include food webs of the Chesapeake Bay, USA, the island of St. Martin, a UK grassland, and a Florida seagrass community, which appear to be the most trophically complete available in the primary literature due to their inclusion of autotrophs and empirically derived estimates of the relative energetic contributions of each trophic link. We show that most (54%) of the 212 species in the four food webs can be unambiguously assigned to a discrete trophic level. Omnivory among the remaining species appears to be quite limited as judged by the standard deviation of omnivores’ energy-weighted food-chain lengths. This allows simple algorithms based on binary food webs without energetic details to yield surprisingly accurate estimates of species’ trophic and omnivory levels. While maximum trophic levels may plausibly exceed historically asserted limits, our analyses contradict both recent empirical claims that these limits are exceeded and recent theoretical claims that rampant omnivory eliminates the scientific utility of the trophic level concept. Williams and Martinez Page 3 The study of food chains and the trophic structure of ecosystems have long been central to ecology (Elton 1927, Lawton 1989, 1995, Wilbur 1997, Post 2002a). Food chains depict the paths through a food web that organic energy travels beginning with basal species and ending with assimilation by a species of interest. A species’ trophic level (TL) indicates the number of times chemical energy is transformed from a consumer’s diet into a consumer’s biomass along the food chains that lead to the species. Convention holds that species that eat no other organisms are basal species with TL=1 while their direct and indirect consumers are at higher levels. Research on TL focuses on patterns common to all ecological systems (Elton 1927, Lindeman 1942, Lawton 1989, 1995, Pimm and Lawton 1978, Pimm 1980, Pimm et al. 1991, Cousins 1987, Yodzis 1989, Martinez and Lawton 1995), patterns that distinguish types of systems (Hairston et al. 1960, Ehrlich and Birch 1967, Briand and Cohen 1987, Moore et al. 1989, Hairston and Hairston 1993, 1997, Polis and Strong 1996, Post et al. 2000), and patterns that distinguish species’ roles within ecological systems (Power 1990, Cabana and Rasmussen 1994, Brett and Goldman 1997, Pace et al. 1999, Vander Zanden and Rasmussen 1996, Vander Zanden et al. 1999, Schmitz et al. 2000) including the role of human exploitation in marine ecosystems (Pauly et al. 1998a, 2002). 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Measuring TL is central to this wide range of trophic ecology (Post 2002a). Food-web research plays a prominent role in measuring TL on a species-by-species and whole-system basis (e.g. Pimm et al. 1991, Polis and Winemiller 1996, Vander Zanden and Rasmussen 1996, Post 2002a). Food webs or, “who eats whom” within ecological systems, describe the food chains in these systems. When the food web includes empirical estimates of energy flows through trophic links, “flow-based TL” is measured by computing food-chain lengths and the relative energetic contributions through chains of different lengths (Levine 1980, Adams et al. 1983). Food webs Williams and Martinez Page 4 usually lack such flow estimates and more simply characterize flows or “links” between species as present or absent. In this binary situation, various measures of consumers’ food chain lengths have been interpreted as measures of consumers’ TL. Pimm (1980, 1982) preferred modal chain length but also identified the extreme measures, the longest and shortest chain to a basal species. Ecologists who argue that most energy flows through the shortest chain to a basal species (e.g., Yodzis 1984, Hairston and Hairston 1993) prefer the shorter extreme while others (e.g., Martinez 1991, Polis 1991, Fussman and Heber 2002) preferred an intermediate measure we call “chainaveraged TL” in which the contribution of each food chain is weighted equally 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Despite the need for webs with link flow information to address central ecological questions, few published species-rich food webs include estimates of energy flows. In contrast, binary food webs are more abundant and tend to have many more species (Cohen et al. 1990, Williams and Martinez 2000). Adding estimates of energy flows through each link in a binary food web requires a great increase in observation effort (e.g., Cohen et al. 1993, Martinez et al. 1999). This additional effort is unnecessary for establishing the TL of basal species and nonomnivorous consumers whose chains to basal species are all of equal length. However, 16% to 77% of the 625 consumer species in the nine food webs studied here have chains to basal species of various lengths (Table 1). Such species include omnivores and their direct and indirect consumers. An accurate measure of TL for these species based on binary webs and independent of energetic flow estimates could significantly increase the scientific productivity of trophic ecologists. In four food webs, we compare six estimates of TL based only on binary link information to the “flow-based TL” based on information that quantifies the energy flow through the webs. We studied five additional binary food webs to further evaluate omnivory, limits to TL, and the Williams and Martinez Page 5 differences between the estimates based on binary links. Our objectives are to compare recently available trophic data to general theories about TL and omnivory while also developing improved approaches for similar endeavors in the future. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 METHODS Data and Terminology We analyzed four of the largest food webs in the primary literature (Tab. 1a) that include 1) relatively many species, 2) empirically derived estimates of the energy flowing through each of the food webs’ links and 3) reasonably resolved basal and other trophic levels within the habitat described. One describes the food web of the Chesapeake Bay, USA (Baird and Ulanowicz 1989). It focuses on the pelagic portion of the Bay emphasizing larger fishes. Another describes an Anolis centered food web on the Caribbean Island of St. Martin (Goldwasser and Roughgarden 1993). The third food web is a UK grassland food web based on endophytic insects found inside the stems of 10 co-occurring grasses (Dawah et al. 1995, Martinez et al. 1999). The fourth is a macroinvertebrate and fish dominated food web of a Florida seagrass community (Christian and Luczkovich 1999). We also studied the following five large, high quality, binary food webs that lack estimates of energy flows (Tab. 1b). The Skipwith Pond food web (Warren 1989) is a speciose freshwater invertebrate web. The food web from the Ythan Estuary (Hall and Raffaelli 1991) emphasizes birds and fish among invertebrates and primary producers. The food web from the Coachella desert (Polis 1991) is a highly aggregated terrestrial web that is also highly connected. The Little Rock Lake food web (Martinez 1991) is a very large and highly resolved food web that includes both pelagic and benthic species. Among a prominent set of 50 Adirondack lake food webs that include only pelagic species (Haven 1992, Martinez 1993), we selected the Williams and Martinez Page 6 largest, the Bridge Brook Lake web. 1 2 3 4 5 6 7 8 9 10 11 12 13