A spatially implicit model fails to predict the structure of spatially explicit metacommunities under high dispersal

Metacommunities are the product of species dispersal and topology. Metacommunity studies often use spatially implicit models, implemented by fully connected topologies, in which the precise spatial arrangement of habitat patches is not specified. Few studies use spatially explicit models, even though real-world metacommunities are likely structured by topology. Here, we test whether a spatially implicit resource consumption model based on a fully connected topology could predict the structure of spatially explicit metacommunities. Having controlled for environmental heterogeneity, we focus specifically on the effects of species dispersal and topology on metacommunity structure. We classified the topologies according to the shortest path between the most distant nodes (i.e. the graph diameter). Topologies with small diameters are tightly connected, whereas large diameter graphs are loosely connected. Some general trends emerged with increasing dispersal rate, such as a hump-shaped pattern in α-diversity, and a plateau followed by a decline in γ- diversity. However, the importance of topology was also apparent: α-diversity peaked at low dispersal rates in small diameter topologies, but at high dispersal rates in large diameter topologies. At low dispersal rates, α-diversity was higher in spatially implicit than in spatially explicit metacommunities. At medium dispersal we detected stronger species sorting in the small diameter than in the large diameter topologies. Increasing dispersal caused α-diversity to decline more dramatically in small diameter topologies. Smaller metacommunities were dominated by regional competitors, whereas larger communities exhibited patterns of species biomass distribution leading to emergent niche structures. Increasing dispersal caused the mean productivity of each patch to undergo partial declines in spatially implicit metacommunities but continue to decline sharply in spatially explicit metacommunities. We conclude that spatially implicit models should be used cautiously when predicting the biodiversity, community composition or ecosystem functions of spatially explicit metacommunities at medium, and especially at high dispersal rates.