Environment complexity and nativeness control the level of coarse-graining at which community assembly is reproducible

Experiments demonstrate that microbial communities sampled from nature and reassembled in laboratory replicates exhibit reproducibility at a coarse taxonomic level (family), yet a high variability at finer scales (species). As was recently shown, the former is ensured by metabolic constraints, while the latter is allowed by functional degeneracy within a family, and multistability of population dynamics. Is the family-level classification special (due to universality of metabolic rules), or should we expect the level at which reproducibility is observed to depend on experimental conditions? What are the theoretical expectations, and what experimentally accessible control parameters could allow predictions to be tested? To address such questions, we built a theoretical framework where a microscopic "ground truth" based on consumer-resource dynamics can be systematically compared to a hierarchy of coarse-grained descriptions. Here, we investigate how the level of coarse-graining at which replicate communities are reproducible is controlled by both environment complexity and environment "nativeness" - the extent to which the test environment resembles the natural context from which the community is originally derived.