Applying Complex Network Theory to Investigate the Fate of Creeping Landslides

Due to climate change and the resulting increase in heavy precipitation, hillsides are predicted to more frequently  transition from slow creep to catastrophic failure. While the pre-failure deformation is sometimes apparent in retrospect, it remains challenging to predict the sudden transition from gradual deformation to runaway acceleration. We apply methods developed to describe the physics of complex systems to investigate the spatiotemporal patterns of slow deformation at active landslides sites, including one that has recently undergone catastrophic failure. We transform measurements of the study sites, such as ground surface displacement, soil moisture, and topographic slope, into a spatially-embedded network in which the nodes are patches of ground and the edges that connect them. We focus primarily on weighting the edges using ground deformation time-series from satellite interferometric synthetic aperture radar (InSAR) data. To eliminate directional or geographical bias when sampling the area, we create a disordered mesh of nodes with Poisson sampling and use Delaunay triangulation to join nodes to their nearest neighbors. This spatially-embedded network is represented as a multilayer network where each layer represents a time slice captured from InSAR. We use community detection, which identifies strongly-correlated clusters of nodes, to identify patterns of instability. We test a variety of network metrics, such as the strength and flexibility of the communities, to quantify patterns of ground deformation leading up to failure. In our preliminary analysis, graphs of several such community metrics (e.g. consistent partitioning into communities, number of communities) show a quiescent period that ends in the weeks immediately prior to failure. These methods therefore show promise as a possible technique for highlighting regions at risk of catastrophic failure.