Colliding Cascades Models for Earthquake Prediction

Wednesday, September 26, 2001 - 9:30am - 10:30am
Keller 3-180
Vladimir Keilis-Borok (University of California, Los Angeles)
1. Colliding cascade (CC) models have been recently introduced to describe development of critical transitions (i.e. abrupt overall changes) in hierarchical non-linear (complex) systems. The models have branching hierarchical structure. The load is applied at the top of the hierarchy and transferred downwards, forming direct cascades. Failures are initiated at the lowest level of hierarchy, and propagate upwards, forming inverse cascades. Direct and inverse cascades collide and interact: loading triggers the failures, failures release and redistribute the load.

2. Three kinds of CC model are developed, different in representation of the loading (differential equations vs. pure cellular automaton) and of interaction between the elements: the interactions are either defined directly , or (according to the concept of Boolean delays equations) replaced by time delays between consecutive switching of the state of an element (loaded vs. unloaded; broken vs. intact).

3. In applications to seismicity loading imitates the impact of tectonic forces, and failures imitated the earthquakes; a major earthquake is the failure at the top level of hierarchy. The models reproduce major heuristic constraints, that is basic features of dynamics of seismicity: seismic cycles, magnitude distribution, clustering, and long-range correlations.

4. The CC models reproduce also (for the first time) the wide variety of seismicity patterns premonitory to strong earthquakes. Moreover, premonitory rise of earthquakes' correlation range has been found on such models first, and then - on observations.

The talk summarizes the recent joint studies by A. Gabrielov (Purdue University), M. Ghil (UCLA), V. Keilis-Borok (UCLA&Russian Ac. Sci), W. Newman (UCLA), D. Turcotte (Cornell U), and I. Zaliapin (.(UCLA&Russian Ac. Sci).