Trends of Global Seismic Noise Properties in Connection to Irregularity of Earth’s Rotation

Abstract

The properties of continuous records of low-frequency seismic noise on a global network consisting of 229 broadband seismic stations located around the world are considered. Changes in the properties of seismic noise, estimated daily for the time interval from the beginning of 1997 to the end of February 2019, are investigated. We consider the generalized Hurst exponent, the singularity spectrum support width, the entropy of the wavelet coefficients, and wavelet-based Donoho–Johnstone index. For the centers of 50 clusters of seismic stations, the average values of these 4 statistics from the 5 nearest operational stations are calculated daily. As a result, 4 multidimensional time series with a dimension of 50 are obtained with a time step of 1 day for more than 22 years of measurements. Average daily values of the noise properties studied, calculated over all cluster centers, have piecewise linear trends, the break point of which is estimated by the principal component method as mid-2003. After the break point, the average values of generalized Hurst exponent, singularity spectrum support width and Donoho–Johnstone index the parameters decrease whereas the entropy increases. This is interpreted as a simplification of the noise structure which is typical for areas of high seismic hazard. Trends in average noise properties after 2003 are accompanied by a linear increase with imposed 3-years quasi-periodic fluctuations in the average value of pairwise correlation coefficients between the values in cluster centers when evaluated in a sliding time window with a length of 1 year. It is hypothesized that the simultaneous simplification of the structure of global seismic noise, an increase in its spatial correlation and an increase in the intensity of the strongest earthquakes in the world after the end of 2004 is a single process associated with the irregularity of the Earth’s rotation. To confirm this hypothesis, a change in the coherence spectrum between the first principal component of the seismic noise properties and the time series of the length of day is estimated.