Abstract : Recent technological advances and long-term data studies provide interaction data that can be modelled through dynamic networks, i.e a sequence of different snapshots of an evolving ecological network. Most often time is the parameter along which these networks evolve but any other one-dimensional gradient (temperature, altitude, depth, humidity, . . . ) could be considered.
Here we propose a statistical tool to analyse the underlying structure of these networks and follow its evolution dynamics (either in time or any other one-dimensional factor). It consists in extracting the main features of these networks and summarise them into a high-level view.
We analyse a dynamic animal contact network and a seasonal food web and in both cases we show that our approach allows for the identification of a backbone organisation as well as interesting temporal variations at the individual level.
Our method, implemented into the R package dynsbm, can handle the largest ecological datasets and is a versatile and promising tool for ecologists that study dynamic interactions.