Oikos 1997

FORUM FORUM FORUM

FORUM is intended for new ideas or new ways of interpreting existing information. It provides a chance for suggesting hypotheses and for challenging current thinking on ecological issues. A lighter prose, designed to attract readers, will be permitted. Formal research reports, albeit short, will not be accepted, and all contributions should be concise with a relatively short list of references. A summary is not required.

Trophic uncertainty vs parsimony in food web research Debal Deb, WWF-India, Eastern Region, Tata Centre 5th floor, 43 Chowringhee Rd, Calcutta 700071, India.

Gut content analysis (GCA) is the most widely accepted method for generalising about a species' food habits. GCA is valuable if the purpose of the study is to determine the frequency or strength of interactions between species, or to establish new food links. However, determining all food links through GCA is impossible for large speciose webs. Furthermore, GCA may not reveal the true nature of linkage dynamics due to environmental and physiological stochasticity. It is therefore parsimonious to assume that linkages between species recorded in the literature will be found in all food webs, if the same prey and predator species occur in those systems. To reveal new linkages, fresh GCA is desirable, but impracticable for large speciose webs containing many rare and endangered species, in which case it may be replaced by several non-dissective methods. High-resolution data for tropical webs could be generated through observations made by trained indigenous peoples.

tions. This, combined with the fact that different nondissective methods have been employed by food web researchers only in recent years (e.g. Havens et al. 1996), seems to indicate that GCA has been held by many authorities to be the most reliable method. Schoenly and Cohen (1991), for example, insist that GCA is indispensable for establishing food linkages, and ought to be conducted afresh for every new description of a food web, in order to ascertain the trophic links actually existing in the system during each period of observation.

Trophic Gut content analysis (GCA) is performed to find out what an animal has eaten, and the finding is subsequently generalized: what has been found in a specimen's stomach would represent the diet items of the species. This kind of inductive generalization is a powerful tool in biology. However, inferring from GCA may not always reveal the true nature of interaction dynamics (Stoner and Zimmerman 1988). For example, the variety and density of prey, access to the food items, predator hunger and gustatory preferences may affect the inferences drawn from the typological GCA. Not only are the trophic links always in a state of flux (Lane 1985), but the directionality of the links also varies according to developmental stages of organisms (Warren 1989, Deb 1995). Thus, the gut contents of today's samples are likely to differ from those of another time; similar samples from different communities may also yield different GCA results, due to their different species compositions and abundances, and also perhaps due to different environmental influences. Considering these and other limitations of GCA, many researchers have adopted several alternative methods, but their explanations as to why GCA was not conducted (Stoner and Zimmerman 1988, Havens 1991, Polis 1991, Deb 1995) often appear to reveal that peer review of such works demanded those explanaOIKOS

78:1 (1997)

indeterminateness

Precision of inferences regarding the structure of a food web crucially depends on the resolution of data. To describe an entire web requires identifying all taxa in the system, which appears too ambitious to accomplish. The checklist of species in a moderately sized community tends to lengthen with time and effort spent on identifying them (Cohen et al. 1993, Havens et al. 1996). As a corollary, the diet spectrum of anyone species is likely to increase indefinitely with efforts to discover them (Polis 1991). One might call this "trophic indeterminateness", which seems to be corroborated by a growing body of evidence. Polis (1991) derived his "species-effort curve" from his study of desert arthropods, and aquatic systems have also yielded similar results. For example, the water flea Daphnia, known to be purely herbivorous, were reported by Gilbert and coworkers (Burns and Gilbert 1986, Gilbert and MacIsaac 1989) to kill and consume small rotifers such as Keratella in the process of filtering algae. Calanoid copepods, known as pelagic herbivores, are now reported also to eat small brachionid rotifers (Warren and Lawton 1987). Until recently, phagotrophic uptake of bacteria by phytoflagellates (Tranvik et al. 1989) was also unknown. Aquatic micro-organisms await intensive studies to reveal further det.ails of feeding behaviour. Terrestrial examples include such common mammalian herbivores 191

as sheep and red deer who occasionally consume

live birds

-

1989), and the Arabian wolf (Canis lupus arabs) in the Sarawat mountain range who thrive mainly on fish and birds, as reported in the March 1995 issue of BBC Wildlife reports (p. 11). Thus, the number of potential food items of an animal may be larger than what is believed to be its actual diet spectrum. By contrast, a potential food organism may not be eaten by the predator, if the prey has anti-predator adaptations: for example, Keratella slacki, Brachionus calyciflorus, and Polyarthra spp. most effectively evade predation from the predatory rotifer Asplanchna (Kerfoot and Sih 1987, Gilbert and Kirk 1988). In spite of this expanded knowledge of feeding biologies, the presence or absence of the trophic links may not always be detectable by random gut examinations due to stochastic environmental and physiological reasons. Predator hunger, for instance, regulates the planktonic clearance rates of both usually-resistant and susceptible prey (Stemberger 1985). Furthermore, a fresh GCA may reveal that a particular resource type is absent from the gut of an animal, but that may not indicate the "absence of a link. Rather, it simply means that the predator did not eat the prey type during the period equal to the gut passage time", when it was examined (Havens 1991). Such spatio-tempora1 uncertainty about the constancy of dietary links may debilitate the very objective of GCA. Schoener (1989) opined that the decision to draw potential links must be either "vetted by experiments ... or by extensive comparative observations, both hard to come by and unlikely to be obtained for most food webs in even the far future, however desirable" (p. 1586). Schoenly and Cohen (1991), however, not only demand such experiments and repeated observations be performed, but would also like to see the ideal student to report the fluctuations of population densities, as well as seasonal changes of selected abiotic environmental parameters. that might give clues to the flux of dietary links. Thus, they suggest that trophic linkages be 4educed from all empirical data in every particular case, instead of inducing them from a few cases. This stand is characteristic of extreme empiricism, and leaves no scope for generalization. One may also argue that Schoenly and Cohen's recommendation implies that first we describe a phenomenon (here, presence/absence of trophic link), and then seek a suitable cause (e.g. _environmental parameters) by reference to which we

can explainthe phenomenon under that description - a procedure that corrupts the scientific method.

Application of Occam's razor In contrast with their own recommendation, Schoenly and Cohen (1991) themselves relied on the published 192

data of food habits of the species comprising their

to meet mineral deficiencyin food (Bazely time-specific ~ebs, constructed "by assuming that a feeding link from prey A to predator B was present if and only if such a link was present in the cumulative web and species A and B occurred together at the time of observation". A cumulative web is a web which incorporates all known trophic links between pairs of species across time (links during summer, winter, day and night, for example). Thus A is assumed to be linked always to B, regardless of how often A eats B, or whether this link is stronger or weaker than other such links in the community. This parsimony of assumption that the links between pairs of (onto)species observed in the past will be observed in the future and in similar systems involves application of Occam's razor, which is essential for all rational enquiry (Lindh 1993). Based. essentially on this argument, most recent food web studies are literature-dependent, and yet show improved generalisations about web statistics (e.g. Martinez 1991, 1992, Havens 1992, 1993, Deb 1995). The essential improvement in these studies has been due to finer resolution of known data rather than using new dietary information. One problem with cumulative webs is that if ontogenetic diet shifts occurred, the simultaneous presence of A and B in their webs would not necessarily signify a feeding relation between them (Schoenly and Cohen 1991). Thus, if B is eaten by the juvenile, but not the adult of species A, the calculation of food links between them whenever A and B co-occur would simply overestimate the number of actual links in cumulative webs. This overestimation could be avoided if the relevant life-history stages of organisms are described separately as distinct web elements in the cumulative webs (Havens 1992, Deb 1995). I would like to call such elements ontospecies (Deb 1995). Thus, species A may be resolved into onto species X (the young of species A) and Y (adult A), such that B (in the above example) is linked with X, but not with y.

Estimating

overestimation

limits

Constructing cumulative webs implies that the organisms"in the web under study use (or, will use) all trophic biologies they are known to use. As a result, the number of food links (L) tends to be overestimated (Pimm et at. 1991). Here the methodological problem is whether the overestimation of L approximates the topologically maximum number of feasible links (Lmax) amongst all onto species, in which case the whole exercise of cumulative web analysis is bound to yield spurious results. Lmaxis defined as

Lmax

= [S(S - If-

I

Si(Si

- I)J/2 = I

SiCS

- sJ/2,

OIKOS 78:1 (1997)

where s; is the number of species on the ith trophic level, and S = L s;. To test if L calculated from cumulative webs might approximate LmaX' I analysed my own data on two freshwater ponds from southern Bengal, and simulated 11 500 randomized webs on computer, using the same pool of onto species (Deb 1995). All trophic links between the onto species were inferred from the literature. The result (Fig. 1) shows that the two estimates move increasingly apart from each other as S increases, and the lower limit of the Lmax range is above the upper limit of the range of L. This indicates that the method of literature-dependence does not overestimate linkages to such extent as to exhaust the possibility of counting all potential links.

Supplementing GCA Information about the presence or absence of potential links between species may serve as fundamental information for constructing food webs, provided that (a) the study concerns itself only with web structure, not with the frequency or strength of interactions (Polis 1991), and (b) a similar assemblage of the same predator and prey species have been studied previously. When a new species is identified in an ecosystem, or when the food habits of a known species are unknown, GCA seems to be necessary for establishing linkages with the other (onto)species in the system. However, 1000

L

500

10

20

30

40

50

5 Fig. 1. The relationship of linkages with system size (S) for II 500 randomized analogs of pond webs, created on rules of non-random linkages inferred from published data. Each dot is co-incident for at least 20 data points. The curves depict power function relationships: (a) L=0.12S21O, and (b) Lmax= 0.17S2.19 OIKOS 78:1 (1997)

estimation of links through GCA would be too arduous and lengthy for large food webs. GCA for the rarer species would 'be especially time consuming. Literature dependence is perhaps the best option for the majority of the rarer taxa whose food links are known (Carney 1995). For new linkage studies, Paine (1988) has acknowledged the various means of estimating food links, other than GCA: direct observation of the predation act, scatology, indirect evidences, immunological study, and "even plausible guesses". All these may supplement published linkage data and even serve as a substitute for fresh GCA, especially when GCA cannot be carried out. Performing GCA for a majority of species in the immensely speciose tropical food webs, most of which exist in the Third WorId, seems impossible to carry out; apart from the unwieldy number of species, the paucity of skilled researchers and/or the rarity and "endangered" status of certain species that should not be sacrificed for GCA, and/or the lack of infrastructure and funds for food web research are the major impediments. In such cases, direct observations of predation act and indirect evidences will be useful (e.g. Lieberman and Lieberman 1987, Dinnerstein and Wemmer 1988). Havens et al. (1996) have used a mixture of direct measurements including bird nestling regurgitant sampling, and inferences from published data. Indigenous peoples in the Third WorId seem to have a wealth of empirical knowledge about the food habits of a large number of species. One effective way of learning about the tropical systems in the Third WorId countries would be to verify the indigenous people's experiences about the food habits of different species and their life history. Of course, one must be cautious in recording such anecdotes, because many of them are arrant superstitions (for example, a popular belief in India is that snakes ingest banana and milk). On the other hand, numerous items of folk knowledge, though easily verifiable, are not even investigated, because they are presumed by many to be scientifically unusable at the best, and grossly unscientific at the worst. In the light of post-modernist critique, this may be explained as a legacy of the Western prejudices against indigenous knowledge base in general. Thus, such facts as the white-breasted kingfisher preying on rats, and the greater adjutant stork consuming crow and domestic chicken are hardly reported in scientific literature. The question as to whether a particular linkage is too infrequent to deserve estimation in food web statistics seems secondary to the problem of high quality food web description, for which identification of as many food links as possible is the primary prerequisite. This could be facilitated by involving an army of local indigenous people trained as "para taxonomists" (NRC 1992) for inventorying local biodiversity as well as for food linkages. The folk knowledge, filtered through trained observation, is likely to supply sufficiently de193

tailed information for high-resolution food web analyses in the future. Acknowledgements - This paper arose from an unknown referee's comment on my previous Gikos paper that literaturedependence might overestimate linkage estimation. I am grateful to R. L. Brahmachary, P. Bhattacharya and S. R. Banerjee for sharing valuable data, and to Heath Carney for perceptive comments on the paper. I also thank Karl Havens and Heath Carney for giving me the opportunity to read their unpublished manuscripts.

References Bazely, D. R. 1989. Carnivorous herbivores: mineral nutrition and the balanced diet. - Trends Ecol. Evol. 4: 155-156. Burns, C. W. and Gilbert, J. J. 1986. Direct observations of the mechanisms of interference between Daphnia and Keratella cochlearis. - Limnol. Oceanogr. 31: 859-866. Carney, H. J. 1995. Food web approaches in biodiversity studies and conservation. - Verh. Int. Verein. Limnol. 26 (in press). Cohen, J. J., Beaver, R. A., Cousins, S. H., DeAngelis, D. L., Goldwasser, L., Heong, K. L., Holt, R. D., Kohn, A. J., Lawton, J. H., Martinez, N., O'Malley, R., Page, L. M., Patten, B. c., Pimm, S. L., Polis, G. A., Rejmanek, M., Schoener, T. W., Schoen1y, K., Sprules, W. G., Teal, J. M., U1anowicz, R. E., Warren, P. H., Wilbur, H. M. and Yodzis, P. 1993. Improving food webs. - Ecology 74: 252-258. Deb, D. 1995. Scale-dependence of food web structures: tropi-

cal ponds as paradigm. - Oikos 72: 245-262. Dinnerstein, E. and Wemmer, C. M. 1988. Fruits Rhinoceros eat: dispersal of Trewia nudiflora (Euphorbiaceae) in lowland Nepal. - Ecology 69: 1768-1774. Gilbert, J. J. and Kirk, K. L. 1988. Escape response of the rotifer Keratella: description, stimulation, fluid dynamics, and ecological significance. - Limnol. Oceanogr. 33: 14401450. - and MacIsaac, H. J. 1989. The susceptibility of Keratella cochlearis to interference from small cladocerans. - Freshwat. BioI. 22: 333-339. HavenS'; K. 1991. Crustacean zooplankton food web structure in lakes of varying acidity. - Can. J. Fish. Aquat. Sci. 48: 1846-1852.

- 1992.Scale and structure in natural food webs. - Science 257: 1107-1109.

194

- 1993. Predator-prey interactions in natural food webs. Oikos 68: Ll7-124. - , Bull, L. A., Warren, G. L., Crisman, T. L., Philips, E. J. and Smith, J. P. 1996. Food web structure in a subtropical lake ecosystem. - Oikos 75: 20-32. Kerfoot, W. C. and Sih, A. 1987. Predation. - Univ. Press of New England, Hanover, NH. Lane, P. A. 1985. A food web approach to mutualism in lake communities. - In: Boucher, D. H. (ed.), The biology of mutualism: ecology and evolution. - Oxford Univ. Press, New York, pp. 344-374. Lieberman, D. and Lieberman, M. 1987. Seeds in elephant dung from Bia National Park. - Biotropica 19: 365-374. Lindh, A. G. 1993. Did Popper solve Hume's problem? Nature 366: 105-106. Martinez, N. D. 1991. Artifacts or attributes? Effects of resolution on Little Rock food web. - Ecol. Monogr. 61: 367-392. - 1992. Constant connectance in community food webs. Am. Nat. 139: 1208-18. National Research Council (NRC) 1992. Conserving biodiversity: a research agenda for development agencies. - National Academy Press, Washington, DC. Paine, R. T. 1988. Food webs: road maps of interactions or grist for theoretical development? - Ecology 69: 16481654. Pimm, S. L., Lawton, J. H. and Cohen, J. E. 1991. Food web patterns and their consequences. - Nature 350: 669-674. Polis, G. 1991. Complex desert food webs: an empirical critique of food web theory. - Am. Nat. 138: 123-155. Schoener, T. W. 1989. Food webs from the small to the large. - Ecology 70: 1559-1589. Schoenly, K. and Cohen, J. E. 1991. Temporal variation in food web structure: 16 empirical cases. - Ecol. Monogr. 61: 267-298. Sternberger, R. S. 1985. Prey selection by the copepod Diacyclops thomasi. - Oecologia 65: 492-497. Stoner, A. W. and Zimmerman, R. J. 1988. Food pathways associated with penaeid shrimps in a mangrove-fringed estuary. - Fish. Bull. 86: 543-551. Tranvik, L. J., Porter, K. G. and Sieburth, J. M. 1989. Occurrence of bacterivory in Cryptomonas: a common freshwater phytoplankton. - Oecologia 78: 473-476. Warren, P. H. 1989. Spatial and temporal variation in the structure of a freshwater food web. - Oikos 55: 209-311. - and Lawton, J. H. 1987. Invertebrate predator-prey body size relationships: an explanation for upper triangular food webs and patterns in food web structure? - Oecologia 74: 231-235.

OIKOS 78:1 (1997)