THE ASSEMBLY OF EXPERIMENTAL WETLAND PLANT-COMMUNITIES

The ability to predict the composition of communities from environmental factors is a central goal of community ecology. We carefully selected a pool of species and subjected it to a range of environmental factors to determine which factors were able to filter out subsets of species. We began with a pool of 20 species and sowed them into 120 wetland microcosms representing 24 different habitat treatments and monitored them for 5 yr. The treatments were fertility, water depth, fluctuations in water depth, soil texture, leaf litter, length of the initial growing season, and invasion by Typha. After 5 yr 14 species persisted; no rare species survived. The experimental communities differed from random expectation and were assembled by rules that constrained their organization. There were strong and consistent effects of fertility, water level, and leaf litter on community composition. Community assembly was modeled as a series of environmental filters. Some aspects of assembly were deterministic: trajectories were constrained within two ''pathway basins'' and species rank abundances were significantly concordant within treatments. Other factors indicated that assembly has a strong stochastic component: 50% of species were present only occasionally and we cannot accurately predict species ranks. Community stochasticity did not show any clear patterns among treatments.