Network analysis of recurrent avalanche events in a 2D granular pile

Abstract

We analyzed the network constructed from the recurrence of an avalanche in a 2D granular pile by comparing its properties to that of a network constructed from the recurrence of events in a random process (null model) [1]. We then infer causal relations in the occurrence of an avalanche when significant deviations from the acausal model is observed. We also examined the “extremal†recurrence (spatially far but close in time) of an avalanche. The distributions of the spatial distances and the in-degree node significantly deviate from the nulll model. From the latter, we infer that the occurrence of an event is highly dependent on the history of all the other events preceding it. Specifically, we found that there is a higher tendency for a suceeding void collapse to happen at the other side of the pile, away from the avalanche event it follows. Non-trivial network properties are also observed indicating the presence of causal structures in the underlying dynamics of void collapse.