Community optimization and ruggedness of ecological landscapes

Many applications require the assembly of an optimal microbial community that maximizes biofuel production, crop yield, or remediation potential. The number of combinations in which these communities can be assembled grows exponentially with the number of species. Therefore, optimization strategies have to rely on heuristic algorithms that iteratively select the best community from a small set of trial communities. The success of such strategies depends on the ruggedness of the landscape of community function. We show that consumer-resource models with and without cross-feeding have unique steady states that depend only on the presence or absence of species in the community. Thus, community selection is a search on an ecological landscape in close analogy with evolution on fitness landscapes in population genetics. We report typical ruggednesses of such landscapes, and discuss how they depend on inter-specific interactions and environmental conditions. We also determine the conditions under which landscape ruggedness can be estimated from incomplete data.