Extended Data Fig. 4: Spatial partitioning by controlled seeding.

a. Partitioning based on controlled seeding. Number of species in local communities (group size) decreases with increasing partitioning. All combinations of the same group size were tested and pooled after growth to determine the richness of the pooled community–the total number of populations. Due to the exponential nature of combination, controlled seeding is difficult to implement in an exhaustive manner for communities with large number of populations. b. Simulations of all 6 major types of interactions in monocultures and cocultures. A check (cross) mark indicates that a population can (cannot) survive by itself. Overall, partitioning promotes coexistence for negative interaction-dominated pairs and impedes that for positive interaction-dominated pairs. The predator-prey interaction shows both trends, depending on whether the populations survive by themselves. Previously published data are consistent with these simulation results. c. Spatial partitioning by controlled seeding in large communities with only negative or positive interactions. Increasing partitioning increases diversity for communities with only negative interactions and decreases for communities with only positive interactions. Each dot represents a randomly generated interaction network, and 10 networks are generated for negative interaction (red trace) and positive interaction (blue trace) networks. The open circles represent the mean across the 10 random interaction networks and error bars represent the standard deviations. d. Robust biphasic dependence is observed with controlled seeding. Partitioning implemented by controlled seeding also generates robust biphasic dependence for communities with both negative and positive interactions. Simulations are done on 10 randomly generated interaction networks of 10 populations with 1:1 ratio of positive versus negative interactions (grey dots). Open circles represent the mean and error bars represent the standard deviation.