Monitoring of seismicity is an important and sensitive tool in studies of seismotectonic processes and earthquake prediction. Recent theoretical and numerical models of seismicity consider seismogenes as evolution of hierarchical, non-linear and dissipative systems and are primarily developed as a new base for large earthquake prediction. This involves an expansion of basic features of seismicity diagnostics by joint analysis of models and phemenology: scaling, similarity, self-similarity, spatio-temporal correlation, response to excitation and predictability at different scales of averaging. These features are used in solving the problems of earthquake prediction through correlation between formalized spatio-temporal and energy parameters of scaling seismicity and location, time and energy of large shocks. These features of real seismicity are poorly known in spite of their fundamental importance, and practically have not been used in the studies of large earthquake prediction in Baikal region.

To assess the predictability of seimicity we investigated energy structure and dynamics of seimicity in the Baikal region and three regions by methods of self-similar processes. We have analyzed the energy structure of seismicity using extrapolation of shock distribution on energy class scale KР by inverse cascade as analog of “Cantorian dust”. Temporal Hausdorff dimension D variations of energy structure in annual samplings of earthquakes with KРі8, recorded in the Baikal region between 1964 and 2002, have been investigated. The basic character of stochastic time process is a Hurst range H. We used the method of index of dispersion of counts (IDC) to calculate it. Assessments of self-similarity of Baikal region seismicity showed that optimization of spatio-temporal data sampling and removal of clusters is a necessary condition to reveal a large earthquake precursor from seismic data.