1691 Ecology, 82(6), 2001, pp. 1691–1702  2001 by the Ecological Society of America PLANT COMMUNITY ASSEMBLY ALONG DENDRITIC NETWORKS OF SMALL FOREST STREAMS OLIVIER HONNAY, 1 WARD VERHAEGHE, 2 AND MARTIN HERMY University of Leuven, Laboratory for Forest, Nature and Landscape Research, Vital Decosterstraat 102, B-3000 Leuven, Belgium Abstract. Although insight into local and regional factors that structure communities is growing, the specific contribution of each of these factors generally remains unknown. Moreover, the extent to which community assembly is intrinsically indeterministic (i.e., the degree to which community composition depends on chance events) remains an open question. The complexity of this issue argues for field studies in a semicontrolled envi- ronment. We studied species richness and community composition of a guild of phreato- phytes along dendritic networks of small forest streams in western Belgium. The streams, situated at the forest-covered headwaters of six different watersheds and characterized by homogeneous environmental conditions, impose spatial dispersal constraints to the plant species. This allowed us to assess the relative importance of deterministic processes guiding community assembly. More specifically, we focused on the role of species migration and colonization processes relative to local processes (competitive exclusion and passive habitat selection) on species richness and on community composition. Our analyses at three con- secutive levels (species number, community composition, and species) revealed that plant community assembly along the stream systems is based on (1) historical and stochastic colonization events of species and (2) subsequent deterministic downstream dispersal pro- cesses. Apparently, due to the influence of historical chance events, deterministic processes do not necessarily result in community convergence. In addition we show that continuous downstream dispersal and colonization can compensate for local extinction processes. This is another piece of evidence that, when studying community composition, it is absolutely necessary to place communities in their spatial contexts. Key words: community assembly; community convergence; dendritic network; Flemish Ardennes, Belgium; hydrochory; mass effect; nestedness; plant colonization; river corridors. INTRODUCTION Community composition has long been considered to be the result of local environmental heterogeneity and local processes such as predation, competition, and disturbance (Ricklefs 1987 and see e.g., Hutchinson 1961, MacArthur 1972, Cody and Diamond 1975, Til- man 1982). Recently, this local determinism has been replaced by a more differentiated view of the factors that generate nonrandom patterns of community com- position. Based on empirical and theoretical evidence, Ricklefs and Schluter (1993) have shown that taxo- nomic components of community organization can be understood only by placing the local community in its historical and spatial context. Specifically, the influ- ences of the available species pool, which is determined by evolutionary and historical processes, and of neigh- boring communities, which interact via species migra- tion, have proven to be important for understanding community composition (e.g., Ricklefs and Schluter Manuscript received 27 September 1999; revised 21 March 2000; accepted 25 April 2000; final version received 19 June 2000. 1 E-mail: olivier.honnay@agr.kuleuven.ac.be 2 Present address: ENVICO b.v.b.a., Dellingstraat 28/B, B-2800 Mechelen, Belgium. 1993, Kadmon 1995, Cornell and Karlson 1996, Caley and Schluter 1997, Pa ¨rtel and Zobel 1998). Although our understanding of factors that structure communities is growing, the specific contribution of each of these factors often remains unknown. More- over, the extent to which community assembly is in- trinsic indeterministic, and hence community compo- sition depends on chance events, remains an open ques- tion (Samuels and Drake 1997). Some authors have asserted that community assembly is deterministic and mechanistic and that the unexplained variance in com- munity composition results from the deterministic ef- fects of historical factors that could not be included in the community analysis (McCune and Allen 1985, Ko- dric-Brown and Brown 1993, Belyea and Lancaster 1999). Although chance and historical effects are unique for each community, studying phenomenolog- ical community response may reveal rules or mecha- nisms behind the community structure (Samuels and Drake 1997). The complexity of the interactions that control com- munity composition reinforces the need for field studies in semicontrolled environments. Following this ap- proach, we studied the composition of forest plant com- munities situated along dendritic networks of small for- est streams fed by base-rich groundwater. The networks