Population fluctuations affect inference in ecological networks of multi-species interactions


K. Wells, Inst. of Experimental Ecology, Univ. of Ulm, Albert-Einstein-Allee 11, DE-89069 Ulm, Germany, and: The Environment Inst., School of Earth and Environmental Sciences, The Univ. of Adelaide, SA 5005, Australia. E-mail: konstans.wells@uni-ulm.de


Local abundance and population fluctuations are key factors affecting the realized interaction frequencies in biotic interactions, but they are commonly ignored when network metrics are calculated over aggregated sets of observations. Here we studied how abundance fluctuations (i.e. demographic and stochastic population dynamics in one of the trophic levels) may affect derived network-level inferences in bipartite ecological networks. Variation at both the species and network level in network indices (d’, Dependence, Fisher's alpha diversity for both levels, H2, weighted NODF) were strongly correlated with the extent of abundance fluctuations, with a strong effect of environmental stochasticity on all indices except NODF; this was the only index for which considerable variation was caused by varying carrying capacities among species. Binary connectance, in turn, does not take interaction frequency (and thus abundance) into account and was only influenced by abundance fluctuations at low population sizes if this led to non-occurrence of ‘true’ interactions.

Overall, abundance and population dynamics are likely to play an important role in determining what is commonly observed and summarized into ecological networks. We suggest that ecological network inference should account for underlying population dynamics and uncertainty in what is observed as interaction frequencies, modelling mechanisms at operative organisational levels below the network rather than using aggregated data of observations. Modelling population dynamics may be a valuable tool in this field to conceptualize and tease apart different sources of variation and express uncertainty in our inference from small samples.