Prediction of Great Japan Earthquake 11 March of 2011 by analysis of microseismic noise. Does Japan approach the next Mega-EQ?

Prediction of Great Japan Earthquake 11 March of 2011 by analysis of microseismic noise. Does Japan approach the next Mega-EQ? Alexey Lyubushin lyubushin@yandex.ru, http://alexeylyubushin.narod.ru/ Institute of Physics of the Earth Russian Academy of Sciences Moscow Ver. 05 May 2018

When and Were the prediction of Tohoku EQ was published before the earthquake Lyubushin A.A. Multi-fractal Properties of Low-Frequency Microseismic Noise in Japan, 1997-2008. - Book of abstracts of 7th General Assembly of the Asian Seismological Commission and Japan Seismological Society, 2008 Fall meeting, Tsukuba, Japan, 24-27 November 2008, p.92. Lyubushin A.A. Synchronization Trends and Rhythms of Multifractal Parameters of the Field of Low-Frequency Microseisms Izvestiya, Physics of the Solid Earth, 2009, Vol. 45, No. 5, pp. 1 394. http://alexeylyubushin.narod.ru/trends_and_rhythms_of_microseisms_synchronization.pdf http://www.springerlink.com/content/u0m8666021127077/ Lyubushin A.A. The Statistics of the Time Segments of Low-Frequency Microseisms: Trends and Synchronization Izvestiya, Physics of the Solid Earth, 2010, Vol., No. 6, pp. 5 554. http://alexeylyubushin.narod.ru/lowfrequency_microseisms_statistics.pdf http://www.springerlink.com/content/dg574m78q2861006/ Lyubushin A.A. Synchronization of multi-fractal parameters of regional and global low-frequency microseisms European Geosciences Union General Assembly 2010, Vienna, 02-07 of May, 2010, Geophysical Research Abstracts, Vol. 12, EGU2010-696, 2010, http://meetingorganizer.copernicus.org/egu2010/egu2010-696.pdf Lyubushin A., Multifractal Parameters of Low-Frequency Microseisms // V. de Rubeis et al. (eds.), Synchronization and Triggering: from Fracture to Earthquake Processes, GeoPlanet: Earth and Planetary Sciences 1, DOI 10.1007/978-3-6-120-9_15, Springer-Verlag Berlin Heidelberg, 2010, 8p., Chapter 15, pp.253-272. http://www.springerlink.com/content/hj2l2115775361/ Lyubushin A.A. Synchronization phenomena of low-frequency microseisms. European Seismological Commission, nd General Assembly, September 06-10, 2010, Montpelier, France. Book of abstracts, p.124, session ES6. http://alexeylyubushin.narod.ru/esc-2010_book_of_abstracts.pdf Lyubushin A.A. Cluster Analysis of Low-Frequency Microseismic Noise Izvestiya, Physics of the Solid Earth, 2011, Vol. 47, No. 6, pp. 488-495 (submitted April 26, 2010) http://alexeylyubushin.narod.ru/cluster_analysis_lowfrequency_microseisms.pdf http://www.springerlink.com/content/bp714871l218287m/ Publications after the Tohoku earthquake and about increasing seismic danger at the Nankai Trough. Lyubushin A.A. (2011) Seismic Catastrophe in Japan on March 11, 2011: Long-Term Prediction on the Basis of Low-Frequency Microseisms Izvestiya, Atmospheric and Oceanic Physics, 2011, Vol., No. 8, pp. 904 921. http://alexeylyubushin.narod.ru/long-term_prediction_tohoku_catastrophe.pdf http://www.springerlink.com/content/kq53j2667024w715/ Lyubushin, A. (2012) Prognostic properties of low-frequency seismic noise. Natural Science, Vol.4,No.8A, 659-666. doi: 10./ns.2012.8087. http://www.scirp.org/journal/paperinformation.aspx?paperid=21656 Lyubushin A.A. (2013) Mapping the Properties of Low-Frequency Microseisms for Seismic Hazard Assessment. Izvestiya, Physics of the Solid Earth, 2013, Vol.49, No.1, pp.9 18. ISSN 1069-3513, DOI: 10.11/S10693513110084. http://alexeylyubushin.narod.ru/seismic_noise_properties_maps_2013_eng.pdf http://link.springer.com/article/10.11%2fs10693513110084 Lyubushin A.A. (2013) Spots of Seismic Danger Extracted by Properties of Low-Frequency Seismic Noise European Geosciences Union General Assembly 2013, Vienna, 07-12 of April, 2013, Geophysical Research Abstracts, Vol. 15, EGU2013-1614, 2013. http://meetingorganizer.copernicus.org/egu2013/egu2013-1614-1.pdf Lyubushin, A. (2013) How soon would the next mega-earthquake occur in Japan? Natural Science, Vol.5, No.8A1, 1-7. doi: 10./ns.2013.58A1001. http://www.scirp.org/journal/paperinformation.aspx?paperid=35770 Lyubushin A.A. (2014) Dynamic estimate of seismic danger based on multifractal properties of low-frequency seismic noise. Natural Hazards, January 2014, Volume 70, Issue 1, pp 471-483. DOI 10.1007/s11069-013-0823-7. http://link.springer.com/article/10.1007%2fs11069-013-0823-7

Asian Seismological Commission, Tsukuba, Japan, 24-27 Nov 2008 The earliest prediction (at the middle of 2008): the magnitude of oncoming catastrophe will be more than 8.3.

The latest prediction: dangerous time for waiting catastrophe begins at the middle of 2010

Positions of 78 seismic stations of broadband network F-net in Japan. N, deg 25.09.2003, M = 8.3 11.03.2011, M = 9.0 128 1 1 1 1 E, deg 148

Main sources of lowfrequency microseisms Atmospheric Pressure Variations Oceanic Waves Low-Frequency Microseismic Noise: Slow Earthquakes, Creep 2 min Periods 1000 min

Multi-fractal singularity spectrum Measure of the random signal variability at the time interval [ t δ/2, t + δ/2 ] Multi-fractal singularity spectrum F (α) and its parameters: α - support width and α - generalized Hurst exponent. 1.0 F (α) - fractal dimensionality of the set of time moments t for which h(t) = α µ(t,δ ) δ h(t) δ 0 0.8 0.6 0.4 0.2 α = α max α min t δ α min α α max 0.0 0.05 0.10 0.15 0.20 0.25 0. 0.35 α

Minimum normalized entropy of squared orthogonal wavelet coefficients distribution Minimum normalized entropy : N En = p log( p ) / log( N) min k = 1 k N 2 2 0 En 1, pk = ck / c j, j= 1 c j k - orthogonal wavelet coefficients, minimum is taken by wavelets from Daubechies family.

Data transform Initial LHZ-records, t = 1 sec t = 1 min Seismic Noise, t = 1 min Downsampling Detrending Donoho share Log(Variance) Kurtosis Estimates within adjacent daily time intervals, t = 1 day α Linear Predictability Index Spectral Exponent α α min Smoothness Index Normalized Entropy

Graphics of seismic noise of the length 1 day with sampling time step 1 minute after removing tidal trends High α Low normalized entropy En Low danger, high variability of stochastic behavior types, a lot of low-frequency spikes because of mutual movements of non-consolidated small blocks of the Earth's crust, energy is not accumulated. Low α High normalized entropy En High danger, the behavior of the seismic noise is much more uniform, small blocks of the Earth's crust are consolidated, the energy could be accumulated. α = 0.249, En = 0.790 α = 0.821, En = 0.685 α = 0.154, En = 0.851 α = 0.757, En = 0.704 α = 0.207, En = 0.855 α = 0.809, En = 0.748 0 200 0 600 800 1000 1200 10 0 200 0 600 800 1000 1200 10 Time, minutes Time, minutes

Maps of multi-fractal singularity spectra support width α. Low α values ( blue and violet regions ) indicate synchronization and danger. 1997.01.01-2003.09.25 α 2003.09.26-2011.03.10 α N, deg 0.58 0.56 0.54 0.52 0.50 N, deg 0.49 0.48 0.47 0. 0.45 0.48 0. 0. 0.43 0. 0. 0. 0.41 128 1 1 1 1 1 1 1 1 1 E, deg 128 1 1 1 1 1 1 1 1 1 E, deg From the beginning of 1997 up to 25 of Sept 2003: the area of future seismic catastrophe is characterized by relatively low α and it is not split into North and South parts. From 26 of Sept 2003 up to 10 of March 2011: the area of future seismic catastrophe is characterized by relatively low α and the previous area of low α values is split into North and South parts. From March 14, 2011 by April, 2018: the North part of the relatively low α values before 25.09.2003 was realized as the area of Great Japan Earthquake 11 of March 2011, M=9.0, whereas the South part is still characterized by low α values.

1997 128 1 1 1 1 1 1 1 1 1 2001 0.51 0.50 0.49 0.48 0.47 0. 0.45 0. 0.43 0. 0.41 0. 0.75 0.70 0.65 0.60 0.55 0.50 0.45 1998 128 1 1 1 1 1 1 1 1 1 2002 0.58 0.56 0.54 0.52 0.50 0.48 0. 0. 0. 0. 0.65 0.60 0.55 0.50 0.45 1999 128 1 1 1 1 1 1 1 1 1 2002.09.24-2003.09.25 0.56 0.54 0.52 0.50 0.48 0. 0. 0. 0. 0. 0.58 0.56 0.54 0.52 0.50 0.48 0. 0. 0. 0. 2000 128 1 1 1 1 1 1 1 1 1 2003.09.26-2004.12.31 0.66 0.64 0.62 0.60 0.58 0.56 0.54 0.52 0.50 0.48 0. 0.50 0.49 0.48 0.47 0. 0.45 0. 0.43 0. 0.41 α The sequence of mean maps of multi-fractal singularity spectra support width of lowfrequency seismic noise waveforms for 16 years of observations at broad-band seismic network F-net at Japan. 128 1 1 1 1 1 1 1 1 1 2005 128 1 1 1 1 1 1 1 1 1 2009 0.51 0.50 0.49 0.48 0.47 0. 0.45 0. 0.43 0. 0.41 0. 128 1 1 1 1 1 1 1 1 1 2006 128 1 1 1 1 1 1 1 1 1 2010.01.01-2011.03.10 0.49 0.48 0.47 0. 0.45 0. 0.43 0. 0.41 0. 128 1 1 1 1 1 1 1 1 1 2007 128 1 1 1 1 1 1 1 1 1 2011.03.14-2011.12.31 0.52 0.50 0.48 0. 0. 0. 128 1 1 1 1 1 1 1 1 1 2008 128 1 1 1 1 1 1 1 1 1 2012 0.51 0.50 0.49 0.48 0.47 0. 0.45 0. 0.43 Small black star hypocenter of earthquake 2003.09.25, M=8.3 Large black star hypocenter of Tohoku megaearthquake 2011.03.11, M=9.0. 0. 0.45 0. 0.43 0. 0.41 0. 0.50 0.49 0.48 0.47 0. 0.45 0. 0.43 0. 0.41 0.58 0.56 0.54 0.52 0.50 0.48 0.52 0.50 0.48 0. 0. 0. 0. 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1

Monthly maps of singularity spectrum support width α, corresponding to 2017

Monthly maps of singularity spectrum support width α, corresponding to 2018

Maps of low-frequency seismic noise wavelet-based normalized entropy, i.e. normalized entropy of the noise waveforms squared wavelet coefficients for the best orthogonal wavelet which is found for each station within each daily time window from the minimum of entropy. 1997.01.01-2003.09.25 En 2003.09.26-2011.03.10 En N, deg 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.65 0.64 0.63 0.62 N, deg 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.65 0.64 128 1 1 1 1 1 1 1 1 1 E, deg 128 1 1 1 1 1 1 1 1 1 E, deg From the beginning of 1997 up to 25 of Sept 2003: the area of future seismic catastrophe is characterized by relatively high values of normalized entropy and it is not split into North and South parts. From 26 of Sept 2003 up to 10 of March 2011: the area of future seismic catastrophe is characterized by relatively high values of normalized entropy and the previous area of high entropy values is split into North and South parts. From March 14, 2011 by April, 2018: the North part of the relatively high values of normalized entropy before 25.09.2003 was realized as the area of Great Japan Earthquake 11 of March 2011, M=9.0, whereas the South part is still characterized by high entropy values.

1997 128 1 1 1 1 1 1 1 1 1 2001 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 2005 2009 0.70 0.69 0.68 0.67 0.66 0.65 0.64 0.63 0.62 0.75 0.70 0.65 0.60 0.74 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.65 0.64 0.63 0.74 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 128 1 1 1 1 1 1 1 1 1 1998 2002 128 1 1 1 1 1 1 1 1 1 2006 128 1 1 1 1 1 1 1 1 1 2010.01.01-2011.03.10 0.74 0.72 0.70 0.68 0.66 0.64 0.62 0.76 0.74 0.72 0.70 0.68 0.66 0.64 0.62 0.60 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.65 1999 128 1 1 1 1 1 1 1 1 1 2002.09.24-2003.09.25 128 1 1 1 1 1 1 1 1 1 2007 128 1 1 1 1 1 1 1 1 1 2011.03.14-2011.12.31 0.72 0.70 0.68 0.66 0.64 0.62 0.76 0.74 0.72 0.70 0.68 0.66 0.64 0.62 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.65 0.64 0.63 0.67 0.66 0.65 0.64 0.63 0.62 0.61 0.60 0.59 0.58 128 1 1 1 1 1 1 1 1 1 2000 2003.09.26-2004.12.31 128 1 1 1 1 1 1 1 1 1 2008 128 1 1 1 1 1 1 1 1 1 2012 0.70 0.69 0.68 0.67 0.66 0.65 0.64 0.63 0.62 0.61 0.60 0.74 0.72 0.70 0.68 0.66 0.64 0.72 0.70 0.68 0.66 0.64 0.62 0.75 0.74 0.73 0.72 0.71 0.70 0.69 0.68 0.67 0.66 0.65 En The sequence of mean maps of minimum normalized entropy of squared orthogonal wavelet coefficients of low-frequency seismic noise waveforms for 16 years of observations at broad-band seismic network F-net at Japan. Small black star hypocenter of earthquake 2003.09.25, M=8.3 Large black star hypocenter of Tohoku megaearthquake 2011.03.11, M=9.0. 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1

Averaged monthly maps of minimum normalized entropy of squared wavelet coefficients, corresponding to 2017

Averaged monthly maps of minimum normalized entropy of squared wavelet coefficients, corresponding to 2018

Maps of low-frequency seismic noise waveforms kurtosis 1997.01.01-2003.09.25 κ 2003.09.26-2011.03.10 κ 55 60 50 55 N, deg 45 N, deg 50 45 35 35 128 1 1 1 1 1 1 1 1 1 E, deg 128 1 1 1 1 1 1 1 1 1 E, deg From the beginning of 1997 up to 25 of Sept 2003. From 26 of Sept 2003 up to 10 of March 2011. From March 14, 2011 by April, 2018. Kurtosis = ( x) ( x) 2 4 2 3, x - increments after removing tidal trends within each daily window

1997 128 1 1 1 1 1 1 1 1 1 2001 128 1 1 1 1 1 1 1 1 1 2005 50 45 35 25 65 60 55 50 45 35 25 70 65 60 55 50 45 35 1998 128 1 1 1 1 1 1 1 1 1 2002 128 1 1 1 1 1 1 1 1 1 2006 50 48 28 60 55 50 45 35 54 52 50 48 1999 128 1 1 1 1 1 1 1 1 1 2002.09.24-2003.09.25 128 1 1 1 1 1 1 1 1 1 2007 60 55 50 45 35 65 60 55 50 45 35 25 60 55 50 45 2000 128 1 1 1 1 1 1 1 1 1 2003.09.26-2004.12.31 128 1 1 1 1 1 1 1 1 1 2008 66 64 62 60 58 56 54 52 50 48 75 70 65 60 55 50 45 35 70 65 60 55 50 45 κ The sequence of mean maps of seismic noise waveforms kurtosis for 16 years of observations at broad-band seismic network F-net at Japan. Small black star hypocenter of earthquake 2003.09.25, M=8.3 Large black star hypocenter of Tohoku megaearthquake 2011.03.11, M=9.0. 128 1 1 1 1 1 1 1 1 1 2009 128 1 1 1 1 1 1 1 1 1 2010.01.01-2011.03.10 128 1 1 1 1 1 1 1 1 1 2011.03.14-2011.12.31 128 1 1 1 1 1 1 1 1 1 2012 55 50 45 35 65 60 55 50 45 35 120 110 100 90 80 70 75 70 65 60 55 50 45 35 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1 128 1 1 1 1 1 1 1 1 1

Averaged monthly maps of seismic noise waveforms kurtosis, corresponding to 2017

Averaged monthly maps of seismic noise waveforms kurtosis, corresponding to 2018

Averaged maps of annual correlation coefficient between multi-fractal singularity spectrum support width α and generalized Hurst exponent α* within moving time window of the length 5 days with mutual shift 7 days. 1997 2003.09.25 2003.09.26 2011.03.10 2011.03.14 2018.04. 1997.0027-2003.7123 N, deg 0.5 0.535 0.5 0.525 0.520 0.515 0.510 0.505 0.500 0.495 0.490 0.485 128 1 1 1 1 1 1 1 1 1 E, deg Before Tohoku EQ the region of future event was extracted by relatively high values of correlation coefficient What is observed for time period March 14, 2011 April, 2018

Averaged maps of annual correlation coefficient between kurtosis and generalized Hurst exponent within moving time window of the length 5 days with mutual shift 7 days. 1997 2003.09.25 2003.09.26 2011.03.10 2011.03.14 2018.04. 1997.0027-2003.7123 2003.7315-2011.1863 N, deg 0.54 0.53 0.52 0.51 0.50 0.49 0.48 0.47 0. 0.45 0. 0.43 0. 0.41 0. N, deg 0.555 0.550 0.545 0.5 0.535 0.5 0.525 0.520 0.515 128 1 1 1 1 1 1 1 1 1 E, deg 128 1 1 1 1 1 1 1 1 1 E, deg Before Tohoku EQ the region of future event was extracted by relatively high values of correlation coefficient. What is observed for time period March 14, 2011 April, 2018

Averaged maps of annual correlation coefficient between minimum wavelet-based normalized entropy and generalized Hurst exponent within moving time window of the length 5 days with mutual shift 7 days. 1997 2003.09.25 2003.09.26 2011.03.10 2011.03.14 2018.04. Before Tohoku EQ the region of future event was extracted by relatively low values of correlation coefficient (high absolute values) What is observed for time period March 14, 2011 April, 2018

Cluster analysis of daily median values of 3 multifractal parameters within moving time window ( t ) * ( t) Let s consider moving time window of the length L =5 days and let ξ = ( α, α, α ) be 3D vector within current time window, t = 1,..., L, t is time ( t ) index, numerating vectors. Our purpose is investigating clustering properties of clouds of 3D vectors ξ with each 1-year time window. In particular, we are interesting what is the best number of clusters. Before making cluster procedure a preliminary operation of normalizing and winzorizing was performed within each time window for each scalar component ( t ) ( ) ξ k of vectors ξ t ( ). Here k = 1,2,3 is the index numerating scalar components of the vector ξ t 1 L ( t ) 2 1 L. Let ξk = L ξ, ( ( t ) 2 t= 1 k σ ) k = ( L 1) ξ t 1 k ξ be sample = ( t ) estimates of mean values and variance of scalar components of the 3D vector ξ. Let s perform iterations which consist in coming to values ( t ) ( t ) ( ) ζ k = ( ξ k ξk ) / σ k and clipping values ζ t k exceeding over and under thresholds ± 4σ k. These iterations are stopped when the values ξ k and σ became stable k and equal to the following values: ξ = k 0, σ 1. After this preliminary operation at each current time window we have a cloud consisting of L 3D vectors ( t) ζ. k = Let s split some cloud into given probe number q of clusters using standard k-means cluster procedure. Let Γ r, r = 1,..., q be clusters, zr ζ / n ζ Γ r mean r ( ) vector of the cluster Γ r, n r be a number of vectors ζ t q within cluster Γ r, n = L. K-means procedure minimizes sum with respect to positions of clusters centers z r. Let the best number of clusters r= 1 r S z z z z1,..., zq min q 2 ( 1,..., q) = ζ r min z1,..., zq r= 1 ζ Γ r J ( q) = min S( z,..., z ) * q was solved from maximum of pseudo-f-statistics: 1 q =. We try probe number of clusters within range 2 q. The problem of selecting PFS( q) = σ ( q) σ ( q) max, σ = J ( q) /( L q) 2 2 2 1 0 2 q 0 σ ( q) ν z z, ν n / L, z ζ / L q L 2 2 ( t) 1 = r r 0 r = r 0 = r= 1 t= 1

Pseudo-F-statistics map as an estimate of current seismic danger (a1)-(a3) plots of daily median values of multifractal singularity spectrum support width α, * generalized Hurst exponent α and minimum Hölder-Lipschitz exponent α min from all 78 stations of broadband seismic network F-net in Japan, green lines present running average within time windows of the length 57 days; (b) plot of the best numbers of clusters for the sequence of clouds consisting of 5 daily 3D * vectors ( α, α, α min ) from moving time window of the length 5 days with mutual shift 3 days. The best number of clusters is defined from the maximum of pseudo-f-statistics. Vertical red line indicates time moment of Tohoku mega-earthquake on March 11, 2011, M = 9.1. Two-dimensional diagram (c) presents dependence of pseudo- F-statistics on the probe number of clusters which is varying from 2 up to within each time window. Plot (d) presents mean value of pseudo-f-statistics averaged by all probe numbers of clusters in dependence on right-hand end of moving time window of the length 5 days. Plot (e) presents the sequence of time moments of strong earthquakes M 7 in the rectangular domain with coordinates 28 ο N Latitude 48 ο N; 128 ο E Longitude 156 ο E which is a rather broad vicinity of Japan islands.

Time Prediction: Squared correlation between mean values of multi-fractal parameters and of microseisms from all F-net stations estimated within 1 year moving time window. 0.8 Strong earthquake is only a foreshock 25.09.2003, M = 8.3 11.03.2011, M = 9.0 0.7 Mean correlation since 2004 0.6 0.5 Best fitted line to the points of minimum correlations 0.4 Blue arrows were plotted at the April of 2010. The 2-nd arrow indicates the middle of 2010 as the lower estimate of catastrophe time moment. 0.3 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 Right-hand end of moving time window of the length 1 year

Dangerous spots of high coherence of GPS noise at Nankai Trough Alexey Lyubushin, Institute of Physics of the Earth, Russian Academy of Sciences, Moscow, e-mail: lyubushin@yandex.ru http://alexeylyubushin.narod.ru/index.htm Network of 13 GPS stations in Japan Map of kernel estimate of probability density functions is presented for nodes of the regular net sized x, which are realized the spatial maximum of the frequency-dependent maximal values of the multiple spectral function of coherence of GPS time series from the nearest 10 workable stations for all 3 components of GPS time series with sampling time step 5 minutes within moving time window of the length 5 days. Time interval 2015-2018. Maximum danger at the vicinity of the point N and 1E http://dx.doi.org/10.131/rg.2.1.2656.66. This result is an independent confirmation of the earlier conclusion made by the analysis of seismic noise http://dx.doi.org/10./ns.2013.58a1001. Data were downloaded from the site of Nevada Geodetic Laboratory: ftp://gneiss.nbmg.unr.edu/rapids_5min/kenv/ 1 0.1 0.01 0.001 0.0001 1E-005 1E-006 1E-007 1E-008 1E-009 1E-010 1E-011 1E-012 1E-013 0.6 0.4 0.2 0 100 1000 Periods, minutes 10 100 1000 Examples of multiple spectral coherence functions in dependence on periods from 10 nearest workable GPS station within time window of the length 5 days. Multiple coherence function: http://dx.doi.org/10.41/ag-3757 http://dx.doi.org/10.11/s106935131069 http://dx.doi.org/10.131/2.1.1815.9049 is based on using canonical coherences: http://alexeylyubushin.narod.ru/synchspect.pdf The spectral matrices were calculated in a moving time window of the length 14 samples (5 days) with mutual shift 288 samples (1 day) for each node of regular grid from 10 nearest workable stations using model of vector autoregression of 5 th order. The GPS station is considered workable in the current time window if its registration interval includes the considered time window and the number of missing values does not exceed a predetermined maximum allowable proportion of the total length equal to 0.1. The missed values are filled using information about records from neighbor time interval of the same length as the length of gaps. Before calculating the spectral matrix in each time window the trend is removed by polynomial of 4 th order and ±3σ winsorizing was performed.

Conclusions. During preparing huge seismic catastrophe geological medium become more simple and homogenous. This manifests in the following properties of lowfrequency microseisms noise waveforms: 1) Multi-fractal singularity spectrum support width α decreases, i.e. the noise structure become more simple, less multi-fractal. 2) Minimum value of normalized entropy of squared orthogonal wavelet coefficients increases. 3) According to the results of seismic noise processing from Japan broadband network F-net the region of Nankai Trough could be the place of the next mega-earthquake. 4) In Japan, since 2003, the regime of natural fluctuations of seismic hazard has been established with a period of about 2 years. The previous 2-year cycle preceded the last strong earthquake Kumamoto on 14.04.2016 in the south of Japan. It follows that the next 2018 will be a year of increased seismic danger for Japan.