Ecology, 83(3), 2002, pp. 703±718 q 2002 by the Ecological Society of America USING STABLE ISOTOPES TO ESTIMATE TROPHIC POSITION: MODELS, METHODS, AND ASSUMPTIONS DAVID M. POST1,2,3 1Department of Ecology and Evolutionary Biology, Corson Hall, Cornell University, Ithaca, New York 14853 USA 2Institute for Ecosystem Studies, Box AB, Millbrook, New York 12545 USA Abstract. The stable isotopes of nitrogen (d15N) and carbon (d13C) provide powerful tools for estimating the trophic positions of and carbon ¯ow to consumers in food webs; however, the isotopic signature of a consumer alone is not generally suf®cient to infer trophic position or carbon source without an appropriate isotopic baseline. In this paper, I develop and discuss methods for generating an isotopic baseline and evaluate the assump- tions required to estimate the trophic position of consumers using stable isotopes in multiple ecosystem studies. I test the ability of two primary consumers, surface-grazing snails and ®lter-feeding mussels, to capture the spatial and temporal variation at the base of aquatic food webs. I ®nd that snails re¯ect the isotopic signature of the base of the littoral food web, mussels re¯ect the isotopic signature of the pelagic food web, and together they provide a good isotopic baseline for estimating trophic position of secondary or higher trophic level consumers in lake ecosystems. Then, using data from 25 north temperate lakes, I evaluate how d15N and d13C of the base of aquatic food webs varies both among lakes and between the littoral and pelagic food webs within lakes. Using data from the literature, I show that the mean trophic fractionation of d15N is 3.4½ (1 SD 5 1½) and of d13Cis 0.4½ (1 SD 5 1.3½), and that both, even though variable, are widely applicable. A sen- sitivity analysis reveals that estimates of trophic position are very sensitive to assumptions about the trophic fractionation of d15N, moderately sensitive to different methods for gen- erating an isotopic baseline, and not sensitive to assumptions about the trophic fractionation of d13C when d13C is used to estimate the proportion of nitrogen in a consumer derived from two sources. Finally, I compare my recommendations for generating an isotopic baseline to an alternative model proposed by M. J. Vander Zanden and J. B. Rasmussen. With an appropriate isotopic baseline and an appreciation of the underlying assumptions and model sensitivity, stable isotopes can help answer some of the most dif®cult questions in food web ecology. Key words: d13C; d15N; isotopic baseline; lake food webs; long-lived consumers; stable isotopes; trophic fractionation; trophic position. INTRODUCTION omnivory that are prevalent in many ecosystems (Paine 1988, Polis and Strong 1996, Persson 1999, Vander There is considerable interest in using stable iso- Zanden and Rasmussen 1999). In contrast, food webs topes, particularly those of nitrogen and carbon, to capture the complexity of trophic interactions in eco- evaluate the structure and dynamics of ecological com- logical communities, but are time-consuming to con- munities (e.g., Peterson and Fry 1987, Kling et al. struct, often subjective in their resolution and scope 1992, France 1995, Vander Zanden et al. 1999, Post et (Paine 1988), and typically hold all trophic links to be al. 2000). One advantage of stable isotope techniques of equal importance, which makes them ineffectual for is that they combine bene®ts of both the trophic-level and food web paradigms in food web ecology. Many tracking energy or mass ¯ow through ecological com- studies use trophic levels because they are simple to munities (Paine 1988, Hairston and Hairston 1993, Po- de®ne, characterize the functional role of organisms, lis and Strong 1996, Persson 1999, Vander Zanden and and facilitate estimates of energy or mass ¯ow through Rasmussen 1999). Stable isotope techniques can pro- ecological communities (e.g., Hairston and Hairston vide a continuous measure of trophic position that in- 1993). The trophic level concept, however, is limited tegrates the assimilation of energy or mass ¯ow through by the strict use of discrete trophic levels and its limited all the different trophic pathways leading to an organ- ability to capture the complex interactions and trophic ism. Stable isotopes have the potential to simultaneous- ly capture complex interactions, including trophic om- nivory, and to track energy or mass ¯ow through eco- Manuscript received 18 September 2000; revised 19 February logical communities (Peterson and Fry 1987, Kling et 2001; accepted 12 March 2001. al. 1992, Cabana and Rasmussen 1996). 3 Present address: National Center for Ecological Analysis and Synthesis, 735 State Street, Suite 300, Santa Barbara, The ratio of stable isotopes of nitrogen (d15N) can California 93101-3351 USA. E-mail: [email protected] be used to estimate trophic position because the d15N 703 704 DAVID M. POST Ecology, Vol. 83, No. 3 of a consumer is typically enriched by 3±4½ relative important to note that there are four unknowns in this 15 15 to its diet (DeNiro and Epstein 1981, Minagawa and equation: trophic position, Dn, d Nsc, and d Nbase. The Wada 1984, Peterson and Fry 1987). In contrast, the trophic fractionation of nitrogen (Dn) is generally as- ratio of carbon isotopes (d13C) changes little as carbon sumed to be between 3½ and 4½ (Peterson and Fry moves through food webs (Rounick and Winterbourn 1987), a critical assumption I evaluate in detail in this 1986, Peterson and Fry 1987, France and Peters 1997) paper. If a robust estimate of Dn is available, solving and, therefore, typically can be used to evaluate the the above equation still requires the quanti®cation of ultimate sources of carbon for an organism when the two of the three remaining unknowns. That is why 15 isotopic signature of the sources are different. In ter- d Nsc is not a suf®cient estimate of trophic position 13 15 restrial ecosystems, d C is often used to differentiate without a good estimate of d Nbase. between diets based on plants with different photosyn- Where consumers acquire nitrogen from more than thetic pathways (e.g., C3 vs. C4; Rounick and Winter- one food web, each with a separate set of primary pro- bourn 1986, Peterson and Fry 1987, O'Leary et al. ducers or detritus sources (e.g., ®sh that feed on both 1992). In lakes, d13C is useful for differentiating be- littoral and pelagic food webs), the model must further 15 tween two major sources of available energy, littoral capture any potential spatial heterogeneity in d Nbase. (near shore) production from attached algae and detri- For a two-source food web, trophic position is calcu- 15 15 tus, and pelagic (open water) production from phyto- lated as: trophic position 5l1(d Nsc 2 [d Nbase1 3 13 15 plankton, because the d C of the base of the littoral a1d Nbase2 3 (1 2a)])/Dn, where a is the proportion food web tends to be enriched in 13C (less negative of nitrogen in the consumer ultimately derived from d13C) relative to the base of the pelagic food web the base of food web one. When the movement of ni- (France 1995). trogen and carbon through the food web is similar, a While it is relatively straightforward to use stable can be estimated using carbon isotopes such that: a5 13 13 13 13 isotope ratios to evaluate food web structure and ma- (d Csc 2d Cbase2)/(d Cbase1 2d Cbase2). This two-end- terial ¯ow within a single system (e.g., Peterson et al. member-mixing model allows for the differentiation 1985, Keough et al. 1996, Hansson et al. 1997), many between two sources, such as the littoral and pelagic critical questions in ecology are best answered through food webs found in lakes, and as written assumes that comparisons across multiple systems (e.g., Kling et al. there is little or no trophic fractionation of carbon and 1992, Post et al. 2000). When comparing among eco- that mixing is linear, assumptions I address further in systems, the d15N and d13C of an organism alone pro- this paper (see Fry and Sherr [1984] and Schwarcz and vides little information about its absolute trophic po- Schoeninger [1991] for discussion and expansion of sition or ultimate source of carbon. This is because this mixing model). Where the number of important there is considerable variation among ecosystems in resources (n)is.2, a minimum of n 2 1 isotope ratios the d15N and d13C of the base of the food web from (or other sources of information) are required to resolve which organisms draw their nitrogen and carbon the system (Fry and Sherr 1984, Peterson et al. 1985, 15 13 (d Nbase, d Cbase; Rounick and Winterbourn 1986, Zo- 1986, Peterson and Fry 1987). For example, Peterson hary et al. 1994, Cabana and Rasmussen 1996, et al. (1986) used two isotope ratios (d13C and d34S) to MacLeod and Barton 1998, Kitchell et al. 1999, Vander evaluate the ¯ow of organic material from three dom- Zanden and Rasmussen 1999; France, in press). With- inant producers (upland plants, marine plankton, and 15 13 out suitable estimates of d Nbase and d Cbase in each the salt marsh plant Spartina) to macroconsumers. 15 13 system, there is no way to determine if variation in the Variability in d Nbase and d Cbase derives from dif- d15N and d13C of an organism re¯ects changes in food ferences in the isotopic ratios of carbon and nitrogen web structure and carbon ¯ow, or just a variation in available for uptake by organisms at the base of the 15 13 the d Nbase and d Cbase. Obtaining the isotopic baseline food web, and through variable expression of fraction- required to estimate trophic position is one of the most ation during uptake.