Splitting of the Double-Frequency Microseismic Peak at Land-Based Seismometers in North America

We performed frequency-dependent polarization analysis for all USArray TA broadband seismic stations from the inception of the project (April, 2004) through the present time. For each hour of data, at each station, the processing resulted in nine frequency-dependent quantities: power on the vertical component, power on the north/south component, power on the east/west component, power of the largest eigenvalue of the spectral covariance matrix, the degree of polarization, the azimuth and inclination of the polarization ellipsoid, the phase difference between the two horizontal components, and the phase difference between the vertical and radial components. Strong seasonal variations in these parameters exist at essentially all stations, however the year-to-year variation is much smaller. Therefore, a single year of data is sufficient to determine the average properties of the ambient seismic wavefield at a specific station.

Here we focus on the characteristics of ambient seismic noise in the double-frequency microseism band of 3-10 s. The strength and dominant period of this energy vary significantly and coherently across North America: proximity to a coastline generally leads to increased amplitude; sedimentary basins show increased amplitudes and shorter dominant periods; and the western U.S. shows longer dominant periods than the central and eastern U.S. Many stations have a bimodal distribution of dominant periods with individual power spectral density estimates that show a splitting of the double-frequency peak into two distinct sub-peaks. Such a phenomenon has been observed previously for data recorded by ocean bottom seismometers, with the shorter-period peak attributed to the local sea and the longer-period peak attributed to more distant, coastally generated microseisms. In the case of the TA data, the splitting appears to arise from simultaneous microseism generation at primarily coastal source areas (Pacific Ocean, Gulf of Mexico, North Atlantic) with distinct, preferred excitation frequencies. Microseism source properties estimated from ocean wave models support this interpretation.