Temporal (Un)correlations Between Coda Q and Seismicity: Multiscale Trend Analysis

This paper introduces a statistical technique, based on the recently developed Multiscale Trend Analysis (MTA), for quantifying correlations between non-stationary processes observed at irregular non-coincident time grids. We apply this technique to studying the temporal correlation between the dynamics of the ductile and brittle layers in the lithosphere. Our results confirm the previously reported strong positive correlation between the coda Q^sup -1^ and seismicity and its drop before major earthquakes observed in California. The proposed technique has significant advantages over the conventional correlation analysis: (1) MTA allows one to work directly with non-coincident time series without preliminary resampling the data; (2) the correlation is defined via the stable objects--trends--rather than noisy individual observations, hence it is highly robust; (3) the correlations are quantified at different time scales. The suggested technique seems promising for the wide range of applied problems dealing with coupled time series.[PUBLICATION ABSTRACT]