Can the metacommunity data matrix predict changes in species incidence and abundance?

Abstract

Metacommunity matrices contain data on species incidence or abundance across sites, compactly portraying community composition and how it varies over sites. We constructed models based on an initial metacommunity matrix of either species incidence or abundance to test whether such data suffice to predict subsequent changes in incidence or abundance at each site. We then tested these models against extensive empirical data on vascular plant incidence and abundance collected from 156 forested sites in both the 1950s and 2000s. Predictions from these models parallel observed changes in species incidence and abundance in two distinctly different forest metacommunities and differ greatly from null model predictions. The abundance model shows greater power than the incidence model reflecting its higher information content. Predictions were more accurate for the more diverse forests of southern Wisconsin which are changing faster in response to succession and fragmentation. Simulations demonstrate that these results are fairly robust to variation in sampling intensity. These models, based only on the metacommunity matrix, do not require data on site conditions or species' characteristics. They thus provide a useful baseline for assessing more complex models incorporating data on species' functional traits, local site conditions, or landscape context. They may also prove useful to conservation biologists seeking to predict local population declines and extinction risks.