COMMENTARY

The ‘‘Goldilocks factor’’ in food webs

Eric L. Berlow*†‡, Ulrich Brose†§, and Neo D. Martinez† *University of California, Merced, Sierra Nevada Research Institute, Yosemite National Park, CA 95389; †Pacific Ecoinformatics and Computational Ecology Laboratory, 1604 McGee Avenue, Berkeley, CA 94703; and §Department of Biology, Darmstadt University of Technology, Schnittspahnstrasse 10, 64287 Darmstadt, Germany

n a well known children’s tale, the harvesting, of one species might cause little girl, Goldilocks, nearly gets dramatic changes in the abundances of herself eaten by bears by boldly other species. Recent work (16–20) has choosing between the three bears’ shown how the body sizes of predators Iporridge bowls: one too hot and another and their prey can determine feeding rates too cold. In this issue of PNAS, the article and biomass flows in food webs. Although ‘‘Size, foraging, and food web structure,’’ large flows of biomass from prey to pred- by Petchey et al. (1), tells a similar story ator may not indicate strong predator con- of food webs in which predator species trol of prey populations (21), such flows choose between prey species that are too can still drive population dynamics (22). large or too small. In doing so, the au- Thus, the coupling of foraging theory with thors show how an easily observed species allometric considerations of metabolic trait—body size—can address long- Fig. 1. Cross-scale integration from individuals to theory may unite the structure of food standing ecological questions about who ecosystems. Petchey et al. (1) model organisms’ webs with biomass flows to predict spe- eats whom in complex natural communi- size-based foraging and integrate it with the net- cies’ population dynamics, which in turn work structure of food webs (red arrow). This work ties. Although previous models explain can elucidate the strengths of dynamic contributes to a broad research program that uses coupling between species (8). In this case, well the general network properties of body size to integrate different levels of trophic food webs, Petchey et al. go further by organization and interactions (blue arrows). species’ influence on each other and on explaining how the particular links found the outcomes of species loss are seen as between species within specific food webs emergent properties of population dynam- can result from optimal foraging of preda- The specific idea that the network ics, driven by foraging behavior and tors on prey that are neither too large nor structure of food webs may emerge from biomass flows that are constrained by spe- too small. Like Goldilocks, optimal preda- optimally foraging individuals is intuitively cies’ body size (15). tors choose food that is ‘‘just right.’’ appealing but technically challenging. Body size also allows integration of the In choosing their food, predators and Petchey et al.’s (1) two mathematical for- overall dynamic persistence of food webs other consumers create food webs—the aging functions require that three or four with the diversity and complexity of food feeding networks among species that sus- parameters be fitted to each food web. web structure (10). The mathematical im- tain the life support systems on earth. Further work is needed to address the probability of large, complex networks Ecologists continue to discover fascinating unexplained variability in these parame- persisting dynamically was recognized regularities in the structure of this interde- ters (up to 26 orders of magnitude) and early on (4). More recent theory finds pendent complexity (2, 3). Elucidating the to reduce the analytical complexity of for- that body size ratios between predators mechanisms responsible for these regulari- mulating and fitting them. Nonetheless, an and their prey occupy a Goldilocks-type ties is a fundamental step in tackling long- important contribution is their linking of ‘‘sweet spot’’ that enables diversity to standing questions in ecology that range body size to prey-handling times within enhance both system-level (9, 16) and from ‘‘What confers stability to complex the foraging functions, both of which fac- subsystem-level (22) stability. These stabi- lizing predator–prey body mass ratios are ecosystems?’’ (4) to ‘‘How does species tors set upper bounds to predators’ prey remarkably consistent with those found in loss alter the abundance of other spe- size. The more successful of the two func- the most comprehensive empirical data cies?’’ (5). Surprisingly simple rule-based tions also sets a lower limit to the body available (23). Buffering by larger organ- models successfully capture the overall sizes of prey species. This concept of con- isms’ low metabolic rate (9, 22) and structure of real food webs (2, 6, 7) and sumers feeding on contiguous ranges of higher mobility among variably productive enable further theoretical exploration by prey body masses within boundaries is patches (24) appear to underlie such sta- stochastically generating webs that mimic deeply rooted in foraging theory (12, 13) bility. Petchey et al. (1) provide more and is highly consistent with observed the overall structure of real webs (6, mechanistic insight into such allometric food web structure (14). However, this 8–10). However, such models lack mecha- integration of link-level biomass flows research has lacked the mechanistic bridge nistic explanations for their input parame- with population dynamics and system-level between foraging theory and food web ters and poorly predict the actual links in stability by showing that these parametric real food webs. Petchey et al. (1) have topology. Petchey et al. provide this bridge sweet spots are consistent with optimal synthesized the allometry of body size by integrating both the foraging conse- foraging. with optimal foraging theory to address quences of body size and the energy con- Leaving dynamics aside, the connection both the first limitation (11), and now the tent of individual prey. between body size and comparative analy- second, by assuming that predators are Between the narrow scale of an individ- more efficient at consuming prey that are ual’s foraging and the broad scale of smaller than themselves, but not too community patterns lie the strengths of Author contributions: E.L.B., U.B., and N.D.M. wrote the small. This ‘‘Goldilocks’’ model of food interactions between populations (15)— paper. choice successfully predicts up to 65% of another area of trophic ecology being in- The authors declare no conflict of interest. the feeding links in real food webs, and in tegrated by the use of body size (Fig. 1). See companion article on page 4191. doing so helps to solve the notoriously Understanding interaction strength in- ‡To whom correspondence should be addressed. E-mail: difficult problem of integrating ecology volves the critical ecological challenge of [email protected]. from individuals to ecosystems (Fig. 1). predicting when the extinction, or over- © 2008 by The National Academy of Sciences of the USA

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800967105 PNAS ͉ March 18, 2008 ͉ vol. 105 ͉ no. 11 ͉ 4079–4080 Downloaded by guest on October 1, 2021 ses of the network structure of ecosystems These failures offer important insights ganism is to subdue and digest. Still, eco- also benefits from the mechanistic detail into the limits of body size alone as an logical interactions involve much more that Petchey et al. (1) provide. Seminal explanatory panacea. Predaceous and than just feeding (e.g., mutualisms, ecosys- early work found that species’ diets within aquatic herbivorous links were, as tem engineering) (38), and feeding in- a food web can form an unbroken seg- expected, much better predicted than par- volves more than just prey body size. But ment, or interval, along a fixed sequence asitic, parasitoid, and pathogenic interac- body size and feeding links seem to be of all species within the food web (25). tions. Similarly, terrestrial herbivorous the ‘‘low-hanging fruit’’ of ecological com- Body size helps explain the hierarchy in interactions were not as well predicted as plexity research. Both are universal and one of the earliest structural food web aquatic ones, and the three nondesert ter- relatively easy to measure for many organ- models, which proposed that larger or- restrial webs were fit rather poorly. All of isms. All species are made up of organ- dered species can eat smaller ordered spe- this suggests that some foraging modes fit isms that have mass, and no species we cies, but not vice versa (26). These two size-based optimal foraging models better know of escapes the food web. As such, ideas were then synthesized by the ‘‘niche than others. Body size may be most im- food webs form an essential core of criti- model’’ that assigned consumers unbroken portant to gape-limited or visual predators cal interdependence created by the class feeding ranges on a niche dimension that and perhaps unimportant to insect herbi- of species interactions for which we tend tends to be located below the consumer vores that choose plants on the basis on to have the most comprehensive data (27). Although this model has been shown leaf chemistry or plant defenses. Similarly, (35). Similarly, body size elegantly links to successfully predict overall food web energy maximization may be balanced by energy content, feeding behavior, and structure (6, 7, 28), Petchey et al. provide avoiding the risk of being consumed (32) metabolic needs. Does this mean that we a testable mechanistic interpretation of or by stoichiometric preferences for par- have chosen to search for our lost coins the niche axis in terms of body mass [see ticular nutrients (33). To better interpret under the lamppost simply because that is also Stouffer et al. (14)]. Such mechanisms such success and failure, it would help to are also relevant to empirical analyses of know whether a random model with fewer where the light is? We don’t think so. food web data that increasingly consider adjustable parameters that simply con- Rather, we may be first picking up the body size (29). Even the issue of para- strains big things to eat small things does coins that are easily visible before crawl- sites—arguably the most numerous of as well, or even better, than Petchey et ing under the car to get the rest. Although metazoan species that turn predator–prey al.’s foraging models. examples of nontrophic interactions and body size relationships on their head These many mutually reinforcing dis- exceptions to foraging based on body size (30)—still supports the strong effect of coveries fuel an increasingly exciting are obvious and abundant, the key unan- body size on the network structure of synthetic research program focused on swered questions are when and under food webs (31). Such empirical analyses body size and trophic ecology (34, 35). what conditions these other factors are have, until now, remained primarily statis- Body size is also key to ‘‘ecology’s big, hot important in predicting food web struc- tical rather than mechanistic in approach. idea’’ (36) that successfully explains a ture and dynamics. Answering these ques- Petchey et al.’s (1) model is interesting broader array of ecological patterns, in- tions requires scientists to tease out the not only in terms of where it fits the data cluding species’ relative abundances, spa- variation in feeding patterns and biomass but also where it fails. Their model suc- tial distributions, and diversity (37). flows that can be simply explained by cessfully predicts a substantial number of Petchey et al. (1) contribute to this broad body size. In doing so, we can help clarify links, but Ͻ50% of the observed links agenda by articulating how body size af- the importance of fascinating, and often were predicted correctly in 12 of the 15 fects not only the energy content of an critical, nontrophic ecological interactions webs. In two webs, Ͻ10% were correct. organism, but also how difficult that or- and processes.

1. Petchey OL, Beckerman AP, Riede JO, Warren PH (2008) 12. Wilson DS (1975) Adequacy of body size as a niche 26. Cohen JE, Briand F, Newman CM (1990) Community Food Size, foraging, and food web structure. Proc Natl Acad difference. Am Nat 109:769–784. Webs: Data and Theory, Lecture Notes in Biomathematics Sci USA 105:4191–4196. 13. Persson L, Leonardsson K, de Roos AM, Gyllenberg M, (Springer, Berlin), Vol 20. 2. Dunne JA (2006) The network structure of food webs. Christensen B (1998) Ontogenetic scaling of foraging 27. Williams RJ, Martinez ND (2000) Simple rules yield com- Ecological Networks: Linking Structure to Dynamics in rates and the dynamics of a size-structured consumer- plex food webs. Nature 404:180–183. Food Webs, eds Pascual M, Dunne JA, Santa Fe Institute resource model. Theor Popul Biol 54:270–293. 28. Dunne JA, Williams RJ, Martinez ND (2004) Network Studies in the Sciences of Complexity (Oxford Univ 14. Stouffer DB, Camacho J, Amaral LAN (2006) A robust structure and robustness of marine food webs. Mar Press, New York), pp 27–86. measure of food web intervality. Proc Natl Acad Sci Ecol Progr Ser 273:291–302. 3. May RM (2006) Network structure and the biology of USA 103:19015–19020. 29. Cohen JE, Jonsson T, Carpenter SR (2003) Ecological populations. Trends Ecol Evol 21:394–399. 15. Berlow EL, et al. (2004) Interaction strengths in food webs: community description using the food web, species 4. May RM (1973) Stability and Complexity in Model Issues and opportunities. J Anim Ecol 73:585–598. abundance, and body size. Proc Natl Acad Sci USA Ecosystems (Princeton Univ Press, Princeton). 16. Emmerson MC, Raffaelli D (2004) Predator-prey body 100:1781–1786. 5. Paine RT (1966) Food web complexity and species di- size, interaction strength and the stability of a real food 30. Lafferty KD, et al. (2008) Parasites in food webs: The versity. Am Nat 100:65–75. web. J Anim Ecol 73:399–409. ultimate missing links. Ecol Lett, in press. 6. Stouffer DB, Camacho J, Jiang W, Amaral LAN (2007) 17. Woodward G, et al. (2005) Body size in ecological 31. Cohen JE, Jonsson T, Muller CB, Godfray HCJ, Savage VM Evidence for the existence of a robust pattern of prey networks. Trends Ecol Evol 20:402–409. (2005) Body sizes of hosts and parasitoids in individual feed- 18. Wootton JT, Emmerson M (2005) Measurement of interac- ing relationships. Proc Natl Acad Sci USA 102:684–689. selection in food webs. Proc R Soc London Ser B tion strength in nature. Annu Rev Ecol Evol Syst 36:419–444. 32. Schmitz OJ (1998) Direct and indirect effects of preda- 274:1931–1940. 19. Bascompte J, Melian CJ, Sala E (2005) Interaction tion and predation risk in old-field interaction webs. 7. Williams RJ, Martinez ND (2008) Success and its limits strength combinations and the overfishing of a marine Am Nat 151:327–342. among structural models of complex food webs. J Anim food web. Proc Natl Acad Sci USA 102:5443–5447. 33. Mayntz D, Raubenheimer D, Salomon M, Toft S, Simp- Ecol, 10.1111/j.1365-2656.2008.01362.x. 20. Reuman DC, Cohen JE (2005) Estimating relative energy son SJ (2005) Nutrient-specific foraging in invertebrate 8. Brose U, Berlow EL, Martinez ND (2005) Scaling up fluxes using the food web, species abundance, and predators. Science 307:111–113. keystone effects from simple to complex ecological body size. Food Webs: From Connectivity to Energetics, 34. Brown JH, Gillooly JF (2003) Ecological food webs: networks. Ecol Lett 8:1317–1325. Advances in Ecological Research, ed Caswell H (Elsevier, High-quality data facilitate theoretical unification. 9. Brose U, Williams RJ, Martinez ND (2006) Allometric San Diego), Vol 36, pp 137–182. Proc Natl Acad Sci USA 100:1467–1468. scaling enhances stability in complex food webs. Ecol 21. Paine RT (1980) Food webs: Linkage, interaction strength, 35. Pascual M, Dunne JA, eds (2006) Ecological Networks: Lett 9:1228–1236. and community infrastructure. J Anim Ecol 49:667–685. Linking Structure to Dynamics in Food Webs, Santa Fe 10. Martinez ND, Williams RJ, Dunne JA (2006) Diversity, 22. Otto SB, Rall BC, Brose U (2007) Allometric degree distribu- Institute Studies in the Sciences of Complexity (Oxford complexity, and persistence in large model ecosystems. tions facilitate food web stability. Nature 450:1226–1230. Univ Press, New York). Ecological Networks: Linking Structure to Dynamics in 23. Brose U, et al. (2006) Consumer-resource body-size relation- 36. Whitfield J (2004) Ecology’s big, hot idea. PLoS Biol Food Webs, eds Pascual M, Dunne JA, Santa Fe Institute ships in natural food webs. Ecology 87:2411–2417. 2:2023–2027. Studies in the Sciences of Complexity (Oxford Univ 24. McCann KS, Rasmussen JB, Umbanhowar J (2005) The 37. Brown JH, Gillooly JF, Allen AP, Savage VM, West GB Press, New York), pp 163–185. dynamics of spatially coupled food webs. Ecol Lett (2004) Toward a metabolic theory of ecology. Ecology 11. Beckerman AP, Petchey OL, Warren PH (2006) Foraging 8:513–523. 85:1771–1789. biology predicts food web complexity. Proc Natl Acad 25. Cohen JE (1978) Food Webs and Niche Space (Princeton 38. Montoya JM, Pimm SL, Sole RV (2006) Ecological net- Sci USA 103:13745–13749. Univ Press, Princeton). works and their fragility. Nature 442:259–264.

4080 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0800967105 Berlow et al. Downloaded by guest on October 1, 2021