Blackwell Science, LtdOxford, UK ERE Ecological Research 0912-38142002 Ecological Society of Japan 174July 2002 503 Food webs in river networks M. E. Power and W. E. Dietrich 10.1046/j.0912-3814.2002.00503.x Review Article451471BEES SGML Ecological Research (2002) 17, 451–471 Food webs in river networks MARY ELEANOR POWER1* AND WILLIAM ERIC DIETRICH2 Departments of 1Integrative Biology and 2Earth and Planetary Science, University of California Berkeley, Berkeley, California 94720, USA Food webs and river drainages are both hierarchical networks and complex adaptive systems. How does living within the second affect the first? Longitudinal gradients in productivity, disturbance regimes and habitat structure down rivers have long interested ecologists, but their effects on food web structure and dynamics are just beginning to be explored. Even less is known about how network structure per se influences river and riparian food webs and their members. We offer some preliminary observations and hypotheses about these interactions, emphasizing observations on upstream–downstream changes in food web structure and controls, and introducing some ideas and predictions about the unexplored question of food web responses to some of the network properties of river drainages. Key words: food chain length; food webs; landscape heterogeneity; river networks; stream ecosystems. INTRODUCTION consumer has access. Ecologists have long pon- dered the historical, environmental and biological Food webs are well described as complex adaptive controls that determine path and chain lengths systems (as lucidly reviewed by Levin (1999)). Like and the impacts of particular web members other complex adaptive systems, food webs have (Hairston et al. 1960; Pimm 1979, 1982; Paine diverse components, linked by flows and (often 1980, 1988; Power 1992a; Power et al. 1996; Post non-linear) interactions, ‘which determine and are et al. 2000). This debate has been somewhat con- reinforced by hierarchical organization of these fused when energy flow paths and population con- components (Levin 1999; p. 12)’. Paine (1980) trol chains were not distinguished. Our present pointed out that two distinct flows create different understanding of these relationships is particularly hierarchies in the same food web. Energy flows limited by our rudimentary appreciation of the from more basal resources up to consumers at spatial and temporal contexts and scales of food higher trophic positions, while ‘top-down’ chains webs. The impacts of web members on each other, of population control link consumers to the and the degree to which energy flow predicts inter- resource populations they regulate or limit, if these action strength, depend largely on which spatial consumers are not suppressed by their own preda- sources of energy and nutrients sustain particular tors. Energy flow paths and population control web members, and how resident versus transient chains are related, but not identical. Organisms these members are in their communities. Synthetic should have greater impact in food webs if they studies that link food web dynamics to spatial have access to better, more productive, or more fluxes of energy, matter, organisms and informa- widely distributed energy sources. Conversely, tion across heterogeneous landscapes (Polis et al. interactions and impacts of other species may 1997; Nakano & Murakami 2001) will contribute determine the energy source to which a particular much to our understanding of these issues. In studies of spatial food webs, as of any com- plex system, trade-offs exist between realism and *Author to whom correspondence should be mechanistic understanding on the one hand, and addressed. Email: [email protected] scope and generality on the other. To explore the Received 26 September 2001. effects of landscape heterogeneity on multitrophic Accepted 12 November 2001. level dynamics, ecologists sensibly began with 452 M. E. Power and W. E. Dietrich simplified conceptualizations. These depict fluxes of sediment, water, organic matter or organisms of organisms, energy or materials across habitat from tributaries into mainstems. Longitudinal boundaries (Holt 1984; Polis et al. 1997), over 2- gradients and their impacts on species distribu- D lattices (Oksanen 1990; de Roos et al. 1991), tions have long interested stream ecologists, but or among islands or patches set in an uninhab- upstream–downstream changes in food web inter- itable matrix (chapters in Tilman & Kareiva actions are just beginning to be investigated. Even 1997). Important recent progress towards land- less is known about the network effects per se on scape realism has been made through more organisms and food webs, but these influences may explicit, sometimes experimental, studies of the help explain and predict ‘cumulative watershed ecological effects of specific landscape features, effects’, which are of great current concern in such as boundary permeability (Cadennaso & Pick- watershed management (Li et al. 1994; Dunne ett 2000; Laurance et al. 2001; Cadennaso et al. in et al. 2001). press) or geometry (Fagan et al. 1999; Anderson & In this paper, we discuss classical and more Polis 1999), or seasonal shifts in the relative pro- recent ideas for controls on food chain length. We ductivities of habitats coupled by trophic exchange then explore how such controls may vary at differ- (Nakano & Murakami 2001). However, one gen- ent positions in drainage networks where the eral feature of landscapes has so far received very energy sources, habitat structure and disturbance little attention with respect to its influence on regimes differ in channels and adjacent water- spatial food webs. River drainage networks sculpt sheds. We draw mainly on our own work with all terrestrial landscapes, defining their relief, dis- students and colleagues from rivers in north section and many other aspects of their heteroge- coastal California. A more comprehensive review neity. Channel and valley characteristics influence of the literature relevant to this topic would repay ecologically significant conditions and are partially effort, but is beyond the scope of this paper. predictable from landscape position (Montgomery & Buffington 1993; Sklar & Dietrich 1998). Here, we explore how conditions arising from longitudi- CONTROLS ON LENGTHS OF ENERGY nal and network structure might affect energy flow FLOW PATHS AND FUNCTIONAL and species performances in food webs. FOOD CHAINS IN FOOD WEBS There are many interesting questions to ask about A TALE OF TWO NETWORKS the properties of food webs (e.g. Cohen 1977; Paine 1980, 1988; Schoener 1989). Questions Food web networks are hierarchical. Energy flows about energy path or chain length are particularly up from lower to higher network positions, and informative for guiding investigations of large- consumer control is exerted down some of these scale variation in the distribution of trophic-level paths, but not others. River drainage networks are biomass (Oksanen et al. 1981; Mittelbach et al. also hierarchical, with gravity driving water, sed- 1988; Power 1992a), as well as issues of practical iment, solutes and organic matter from ridges interest. If we want to conserve native species, dividing watershed down into channels, and from sustainably harvest a resource, or suppress biologi- headwaters down mainstems to lowland floodplain cal pests, we need to know the energy flow paths rivers and estuaries. With the exception of winds that support the groups of interest and the top- and local back eddies in turbulent water, only bio- down controls on their abundance. logical fluxes (migrations and other movements of Classical hypotheses for environmental controls individuals) drive materials upstream or upslope. on the length of food chains (which unfortunately Conditions, resource fluxes and biotic interactions do not distinguish energy flow paths from top- experienced by aquatic, riparian and terrestrial down chains) discuss two environmental variables organisms in watersheds vary down longitudinal and one evolutionary factor. gradients from headwaters to lowland habitats (Vannote et al. 1980; Montgomery & Buffington 1 Productivity/efficiency. Chains should lengthen as 1993, 1997). Conditions also vary abruptly, for fluxes of limiting resources or energy to food example, where network confluences inject pulses webs increase, or as consumers increase their Food webs in river networks 453 efficiency of resource capture or conversion cient appendages for dismembering prey (Vogel (Fretwell 1977; Pimm 1979; Oksanen et al. 1981; Power 1987; Power et al. 1997). The rela- 1981; Fretwell 1987; Pimm & Kitching 1987). tionship of food chain length to habitat size may 2 Disturbance/stability. Chains should be shorter be less clear, however, in habitats such as rivers in more frequently disturbed environments where cross-habitat exchanges have strong effects (Pimm & Lawton 1977; Pimm & Kitching (Polis et al. 1997; Nakano et al. 1999). Analyses of 1987). river or adjacent terrestrial food webs may over- 3 Design constraints. Pimm (1979) argues that it look crucial interactions if the boundary of obser- may be impossible for evolution to build a vation or experimentation is drawn at the river Pterodactyl predator, for example, because an surface (Wallace et al. 1997; Nakano et al. 1999; organism large enough to subdue one could not Power & Rainey 2000; Nakano & Murakami fly to catch it. 2001; Sabo & Power in press a, b; Power et al. in press). Despite many food web surveys, particularly Post (unpubl. data, 2001)