Recovery of model coral communities: complex behaviours from interaction of parameters operating at different spatial scales

Because scientists tend to define questions that are commensurate with established and tractable methods of analysis, ecologists concerned with the dynamics of communities and ecosystems have given too little attention to phenomena at large spatial scales and to parameters that have a spatial context. Cellular automata (CA) are a useful tool for study of parameters that have a spatial component and for analysis of system behaviours over a range of spatial scales, and thus for identifying the emergent properties of communities and ecosystems. We illustrate this with simple CA models of a system of coral reefs that includes corals of two types of life-history, viz. broadcast spawning corals which release gametes and brooding species which release larvae that can settle almost immediately. We examine the influence on system recovery after large-scale disturbances of 3 parameters that manifest at different spatial scales, viz. reef density (a parameter affecting large scale connectivity), local hydrodynamic retention of larvae around individual reefs ( meso scale), and the feeding behaviour of crown-of-thorns starfish (small scale). The results indicate that recovery depends on the system level properties of overall connectivity and among-reef variance in coral cover. Also, there are strong interactions among parameters. Conversely, the degree of larval retention is less important to recovery when reef density is high. At high levels of disturbance recovery of broadcasting corals is more sensitive than that of brooders to starfish feeding behaviour, particularly if inter-reef distances are short. There are also thresholds in the qualitative pattern of recovery. The usual pattern of dominance of broadcasting over brooding species can be reversed if poorly connected reefs with low levels of larval retention suffer high levels of disturbance. For either type of life-history there are levels of disturbance from which the system does not recover. The values of these thresholds depend on all parameters we examined, including the life-history type