MODELING THE EFFECT OF TIME AVERAGING ON THE COMPOSITIONAL FIDELITY OF DEATH ASSEMBLAGES

Species composition and diversity of ecological communities vary naturally over time, even in the absence of a secular trend. Because death assemblages mix the dead residue of successive living community states, they are also expected to vary over time. Inherent death assemblage variability may be a significant source of noise observed in quantitative analyses of paleocommunity data and has the potential to obscure biofacies. Assessing how the variability of a real living community and its corresponding death assemblage relate through time requires longitudinal data that are not currently available and are difficult to acquire, so a modeling approach was adopted to address this issue. Local communities were simulated using a neutral model. In each time step, a fixed proportion (μ) of individuals in a living community of size J died and contributed their remains to an accumulating death assemblage. Dead organisms were replaced either by immigration from a metacommunity (whose abundance structure was determined by a diversity parameter, θ) or by birth from local survivors; the ratio of immigration to local birth was determined by a connectivity parameter (m). Loss of individuals from the death assemblage was modeled stochastically using a constant probability of destruction in each time step (λ). Following the living community and death assemblage over 1000s of generations allowed characterization of 1) the similarity between the two at any moment in time and 2) the inherent range of compositional variation of both. The similarity of coeval live and dead assemblages was greatest when the ratio μ/λ equaled ~1.0 and decreased for both μ/λ<1 and μ/λ>1; the degree of decrease was positively correlated to θ (more diverse communities showed greater dissimilarity between live and dead assemblages). The ratio of live-to-dead compositional variation was also strongly related to μ/λ: when μ/λ<1, death assemblages were more variable than live, and when μ/λ>1, live assemblages were more variable than dead. The ratio μ/λ also determines the size of the death assemblage relative to the living community at steady state. Overall, the model supports the conventional wisdom that increased time averaging acts to dampen death assemblage variability, but it can also amplify the discrepancy between live and dead assemblage composition.