The Assembly of Ecological Communities: A Minimalist Approach

1. Communities develop through a process, community assembly, in which species invade, persist, or become extinct. 2. Community assembly is a sequence of different community states. Each state is a unique combination of species' presence or absence. The community transition graph is the directed graph of the transitions between states. Not all the transition graphs are ecologically sensible. 3. We investigated the statistical distribution of different community transition graphs that contain varying numbers of species. Does the species composition of a community persist? Or does it cycle, and, if so, through a few recognizable states or through a complex sequence that might appear superficially random? 4. We employ three recipes to construct the transition graphs. In the random transition graphs, the directions of the transitions are specified entirely at random. In the second recipe, we exclude all cycles by assigning each state a random `height.' The community transition graph moves from a `lower' state to one of its `higher' neighbours. This produces landscape transition graphs. The third recipe adds one ecological constraint to the random assembly models. We remove the simple, ecologically implausible cycles to produce minimal transition graphs. 5. Random transition graphs most commonly have one persistent state. Landscape transition graphs commonly have many more persistent states and the number increases rapidly with the number of species in the system. Persistent cycles are rare in the random graphs and are impossible in the landscape graphs. 6. Persistent states are nearly always reachable from the empty community in the random graphs. In the landscape graphs, with increasing numbers of species, the total number of persistent states increases more quickly than the number of reachable states. 7. Cyclical changes in composition en route to a persistent state are a dominant feature of the random graphs. The path lengths to a persistent state are often very long. 8. The third recipe excludes only the simple cycles. These minimal transition graphs have more than one persistent state. The number of persistent states increases rapidly with the number of species in the model. We did not find persistent cycles. Thus, models with even minimal ecological constraints appear very similar to the landscape transition graphs. Assembly resembles the process of evolution. It looks as though there is something that pushes the assembly towards peaks in a landscape. 9.