Statistics of multiscale fragmentation in the Primorsky fault zone

Tectonic stresses cause rock deformation and disintegration. We examined the fragmentation statistics of brittle rocks composing the damage zone of the Primorsky Fault of the Baikal Rift Zone at scales ranging from microns to kilometers. The fault rocks analyzed include different lithologies and shear-strain magnitudes. We use the convolutional neural network algorithm to automate the mapping of fractures in images and faults in topographic data and statistically test for presence of power, lognormal or Weibull laws. Fault-rock fragmentation obeys lognormal statistics at scales from 10− 6m to 104m, and the shape parameter (σ) is preserved and varies in the range 1.4–2.0. We demonstrate that summarizing the truncated data may lead to compilation artifact and incorrect conclusions about the power law behavior. We proposed a statistical fragmentation model to fit to experimental logarithmically distributed data. At all scales the rate of destruction depends on the fragment size as a power law. Findings should be incorporated in models estimating fault geometry characteristics and evolution of earthquake source.