Ecology, 88(7), 2007, pp. 1696–1706 Ó 2007 by the Ecological Society of America METACOMMUNITY MODELS PREDICT THE LOCAL–REGIONAL SPECIES RICHNESS RELATIONSHIP IN A NATURAL SYSTEM BERNARD HUGUENY, 1,3 HOWARD V. CORNELL, 2 AND SUSAN HARRISON 2 1 IRD, Laboratoire d’ e ´cologie des hydrosyste`mes fluviaux, Universite ´ Lyon1, 43 Boulevard du 11 Novembre 1918, 69622 Villeurbanne Cedex, France 2 Department of Environmental Science and Policy, University of California, Davis, California 95616 USA Abstract. Many natural communities exhibit positive relationships between local and regional species richness (LSR–RSR relationships), which can be either linear or curvilinear. Previous models have shown that the form of this relationship depends on the relative rates of colonization and extinction and the sensitivity of these rates to competition. We use simple models to show that the LSR–RSR relationship also depends on the type of metacommunity structure (Levins-like or mainland–island), and our models generate a wider range of realistic forms than do most previous models. We parameterize and test our models with two independent data sets for Daphnia in rock pools on islands in Finland and Sweden. We find that the Levins-like model with competition correctly predicts the observed LSR–RSR relationship and provides the best fit to the average local species richness per island. Simulations show that our models are robust to relaxing our assumption of identical species properties. Our study is one of the first to make and successfully test quantitative predictions for how a widely studied community pattern, the LSR–RSR relationship, arises from metacommunity dynamics. Key words: colonization; community; competition; Daphnia; extinction; local richness; mainland–island model; metacommunity; regional richness; species richness. INTRODUCTION Regional and historical processes, as well as species interactions within local habitats, can profoundly affect the species diversity of ecological communities (Ricklefs 1987, Cornell and Lawton 1992, Ricklefs and Schluter 1993). As a result, multiple spatial and temporal scales must be considered when asking why some communities support more species than others. This broadening of perspective has been driven in part by the frequent occurrence of positive relationships between local and regional species richness (Cornell and Karlson 1997, Lawton 1999), which imply that local communities are affected by factors that operate at large spatial and temporal scales to shape regional diversity. The inter- pretation of the local–regional species richness (LSR– RSR) relationship has been contentious however, particularly with regard to the importance of interspe- cific competition (Loreau 2000, Shurin et al. 2000, Shurin and Allen 2001, Leibold et al. 2004, Ricklefs 2004, He et al. 2005, Hillebrand 2005, Shurin and Srivastava 2005). Linear relationships have sometimes been taken as evidence that competition is weak or nonexistent at the local scale, while curvilinear (i.e., upward convex) relationships have been interpreted as possible evidence for competition (e.g., Terborgh and Faaborg 1980, Ricklefs 1987, Cornell and Lawton 1992). Inferring processes from patterns can be mislead- ing without an appropriate theoretical background however, and ecologists are still searching for a satisfactory way to link the widely observed linear and curvilinear LSR–RSR relationships to underlying com- munity processes. Caswell and Cohen (1993) were the first to derive the LSR–RSR relationship from what is now termed a ‘‘metacommunity’’ model, in which a region is subdi- vided into a number of habitat patches, and each patch is able to support a local community drawn from the regional species pool (Hastings 1987, Hanski and Gilpin 1991, Wilson 1992, Leibold et al. 2004). In Caswell and Cohen’s (1993) and subsequent models, local richness is the number of species within local patches where species interact. Regional richness is either the number of species from an external source, or the number of species summed over all patches in the metacommunity (Caswell and Cohen 1993, Hugueny and Cornell 2000, Shurin and Allen 2001, Mouquet and Loreau 2003, Mouquet et al. 2003, Fukami 2004, Hillebrand 2005). A key finding of these models is that linear LSR–RSR relationships, which are observed in the majority of real- world cases, do not indicate that competition is weak or absent; linearity can arise from rapid dispersal (Caswell and Cohen 1993, Bell 2003, Mouquet and Loreau 2003, Fukami 2004), disturbance (Caswell and Cohen 1993, Mouquet et al. 2003, Hillebrand 2005), or predation (Shurin and Allen 2001), even in the presence of strong Manuscript received 13 November 2006; accepted 15 November 2006; final version received 5 January 2007. Corresponding Editor: L. Stone. 3 E-mail: hugueny@biomserv.univ-lyon1.fr 1696