1916 Ecology, 86(7), 2005, pp. 1916–1925  2005 by the Ecological Society of America FOOD-WEB TOPOLOGY VARIES WITH SPATIAL SCALE IN A PATCHY ENVIRONMENT ROSS M. THOMPSON 1 AND COLIN R. TOWNSEND Department of Zoology, University of Otago, P.O. Box 56, Dunedin, New Zealand Abstract. The majority of food-web studies currently used to test ecological theory have integrated information over large spatial and temporal scales. We aimed to assess the degree to which food webs display patch-scale variation, and the consequences for emergent properties at the larger scale of the stream reach. Spatial heterogeneity in ecological con- ditions (habitat structure and food resources) and food-web structure were measured in three streams. All food webs were constructed using equivalent effort at a patch scale (0.06 m 2 ) and a reach scale (30-m stream length). A mosaic of habitat structure and food resources was reflected in considerable variability in food-web structure among patches, but there was less variation within than among streams. The variability in food-web attributes among patches could not always be related to ecological conditions, but food resource availability affected connectedness of food webs and trophic structure (measured as functional feeding groups). Of particular note was the result that reach-summary food webs were consistently different from patch-specific food webs in each stream. Reach-scale food webs underes- timated connectance but overestimated prey : predator ratios and the number of trophic links. Summary webs sometimes placed species together that, in fact, did not coexist in the field. Such within-site food-web heterogeneity needs to be taken into account in future multiple-site comparisons of food-web structure. Key words: connectance; food web; ecosystem size; productivity; South Island, New Zealand; spatial scale; species richness; stream patch vs. reach; streams. INTRODUCTION Spatial patchiness of local ecological conditions is a recurrent feature of all ecosystems. Even apparently uniform habitats such as pelagic marine systems dis- play high variability (e.g., Maar et al. 2002), and spatial variability in animal trophic organization has been de- scribed in environments ranging from marine (e.g., Vet- ter 1998, Pakhomov et al. 2004), to terrestrial (e.g., Moore and de Ruiter 1991, Robertson and Freckman 1993), to freshwater (e.g., Stone and Wallace 1998, Garvey et al. 2003). Processes responsible for this pat- terning include variability in physical habitat (e.g., Downes et al. 1993, Yasuhiro et al. 2004), resource supply (e.g., Sinsabaugh et al. 1991, Maar et al. 2002), interspecific competition (e.g., Rader and McArthur 1995, Harrison and Hildrew 2001), parasitism (e.g., Mouritsen and Poulin 2002), and predation (e.g., Zerba and Collins 1992). The incorporation of patch variation into models of food-web structure and dynamics has shown that this characteristic is an important stabilizing influence for the persistence of species (e.g., Pahl- Wostl 1993, Callaway and Hastings 2002, Benedetti- Cecchi 2003). Detailed reviews of the literature (e.g., Manuscript recived 31 August 2004; revised 7 December 2004; accepted 9 December 2004. Corresponding Editor: O. J. Schmitz. 1 Present address: Biodiversity Research Centre, Univer- sity of British Columbia, 6270 University Blvd., Vancouver BC V6T 1Z4 Canada. E-mail: thompson@zoology.ubc.ca Holt 2002) and reanalysis of existing data (Brose et al. 2004) have strongly indicated both the pervasiveness and strong influences of patchiness on local patterns of species coexistence. Despite an awareness of the importance of patch var- iability in patterns of co-occurrence of species, large- scale studies of food webs have assembled connectance webs with little regard for the spatial organization of species. Many of the largest and most highly resolved food webs collected to date (e.g., Cohen 1989, Martinez 1991, Havens 1992, Waide and Reagan 1996) have de- scribed summary food webs on very large scales (e.g., lake, forest, or island scale). In these cases, the scale at which food webs are described is much larger than the scale at which the component organisms operate, obscuring meso- and micro-scale spatial patterns such as local refugia and microhabitat segregation that may preclude some species from coming into contact with one another. This has important consequences for the interpretation of some food-web patterns. Analyses of connectance, for example, interpret the number of tro- phic links that do occur relative to the number that could occur. If spatial patchiness precludes many spe- cies from actually coming into contact, then such anal- yses may be misleading. A number of studies have sought to determine the ecological factors that are correlated with food-web attributes. Meta-analyses of data gathered at relatively coarse spatial scales (e.g., Briand 1983, Briand and Cohen 1987, Lawton 1989) have revealed few gener-