ИСТИНА

The vertical seismic records with sampling rate 1 Hz obtained from global IRIS broad-band seismic network (1988-2009, 131 stations) and Japan F-net (1997-2010, 83 stations) were analyzed by estimating their multi-fractal parameters: the width of singularity spectra support, generalized Hurst's exponent and coefficient of singularity spectra asymmetry. Besides that linear predictability index of waveforms, spectral exponent and logarithm of variance were estimated as well. These statistics were calculated within adjacent "short" time windows of the length 30 minutes for initial 1 Hz data and for time windows of the length 1 day for records after coming to 1 minutes sampling. The seismic stations were split into a number of spatial clusters (7 clusters for global IRIS network and 5 clusters for F-net). The median values of all parameters were taken from stations in each cluster. A multiple correlation measures for different combinations of parameters were estimated within moving time window of the length 1 and 2 years for these median time series with uniform sampling time interval 1 day. Using of long time windows for estimating multiple correlation measure allows to average influence of storms and hurricanes. The sequence of waves of microseisms noise essential global synchronization was extracted. The most strong synchronization effects correspond to time interval 2003-2007 and the 2nd one started at the beginning of 2008 and is continuing till now. The microseisms field at Japan islands transfers to high level synchronization of its parameters starting from the middle of 2002, one year before the Hokkaido earthquake, 25 of September, 2003, M=8.3. This high level of synchronization keeps rather constant up to the current time. Based on the statement of the theory of catastrophes that synchronization is one of the flags of an approaching catastrophe, it may be suggested that the Hokkaido event, notwithstanding its power (M=8.3), could be only a foreshock of a still stronger earthquake at the region of Japan's islands. The cluster analysis of 7 median daily statistics from the whole network indicates a strong linear trend of cluster exponent starting from 2007 which is continuing till now. This trend peculiarity is similar to the trend before 2003 event. The peculiarities of correlation coefficient estimate within 1 year time window between median values of singularity spectra support width and generalized Hurst exponent indicates that starting from July of 2010 Japan islands come to the state of waiting strong earthquake.