The Time Dependence of Ambient Noise Beneath the Deep Sea Floor

R A Stephen*, S T Bolmer*, J A Collins*, J A Hildebrand+, J A Orcutt+, K R Peal*, F N Spiess+, and F L Vernon+

* Woods Hole Oceanographic Institution
+ Scripps Institution of Oceanography


     Between February and May 1998, the Ocean Seismic Network Pilot Experiment (OSNPE) acquired over 115 days of broadband borehole seismic data (0.001-7.5Hz) at ODP Site 843B in 4407m water depth about 225km south of Oahu. The objective of the OSNPE was to compare different styles of seafloor broadband installations in preparation for extending the Global Seismic Network to the deep ocean basins. In this abstract we focus on the time dependence of the ambient noise as acquired on the KS54000 sensor, installed 240???m below the seafloor in the upper igneous basement. The borehole seismometer was developed for deep ocean applications by Scripps IGPP, the autonomous seafloor acquisition system was developed by WHOI, and the wireline re-entry technique was developed and carried out by the Scripps Marine Physical Laboratory. During the OSNPE we acquired seafloor current measurements near the site and sea state and wind data were obtained from near-by NOAA weather buoys. Estimates of the tides at the site are obtained directly from the KS54000 itself at very long periods.

Figure 1
: This true amplitude spectrogram summarizes the vertical component ambient noise in the band 0.6mHz to 7.5Hz for over 110days of the broadband borehole seismometer deployment at OSN-1. The major frequency bands are identified in Figure 4. The diagonal events near 20mHz are the dispersed surface waves from large earthquakes. The steady bands between 0.1 and 0.5Hz correspond to shear modes or reverberations within the sediment column.

Figure 2
: This true amplitude spectrogram summarizes the horizontal component ambient noise similar to Figure 1.

Figure 3:
This vertical component spectrogram has had the mean spectra (see Figure 4) removed in order to emphasize the variability in the spectral level.

Figure 4
: The geometrical mean of all of the spectra in Figure 1 is shown. Also shown are the largest and least spectral values at each frequency. The USGS high and low noise models (land models) are based on a synthesis of data from land and island stations. The labels indicate frequency bands in which the time dependence of ambient noise has similar behavior. The very high levels in bands B through F are arrivals from the large (Mw=7.9) Balleny Islands earthquake on Julian Day 84

Figure 5
: The mean noise level in the infra-gravity band (0.6-3mHz, band A in Figure4) correlates extremely well with tides. The 'tidal' curve is just the time series of the KS54000 vertical component. In the infra-gravity band the borehole sensor exhibited much higher noise levels than a co-located buried sensor, suggesting that the sensor is suffering from instrument or installation noise. The correlation with tides indicates that the noise may be related to fluid-flow in the well or that it may be caused by non-linearities in the sensor. None of the spectral bands shown in Figure 4 correlated with seafloor currents.

Figure 6
: The noise notch minimum (8-55mHz, band C) rises dramatically after most earthquakes observed on the IRIS DMC catalog. The smaller peaks that do not occur at a known earthquake event may be caused by smaller local and regional earthquakes that are not observed on the global network. If we use -175dB as the threshold for an 'event' there are about 34 of these local and regional events. All of the events above -160dB can be associated with quakes in the IRIS DMC catalog.

Figure 7
: The highest levels of the microseism peak (120-350mHz, band G) correlate very well with sea state in the band 60-175mHz as predicted by the wave-wave interaction mechanism.

Figure 8
: Mean noise levels in the HOLU spectrum (0.35-2.0Hz, band H) also correlate with sea state at half the frequency (0.175-1.0Hz).

Figure 9
: As for the ver low frequencies there is a strong correlation between mean noise in the VLF band and tides.


      Between February and May 1998, the Ocean Seismic Network Pilot Experiment (OSNPE) acquired over 115 days of broadband borehole seismic data at ODP Site 843B in 4407m water depth off Oahu. The borehole seismometer was clamped in casing within the upper 6m of igneous basement beneath 242m of sediment. In addition to over fifty earthquake events that were observed, ranging from a 4.5Mb event at 44° epicentral distance to the 7.9Mw Balleny Islands earthquake at 91° epicentral distance, this data set provides an opportunity to study the time dependence of ambient noise in a borehole in the deep sea over a four month duration. The ambient noise behavior falls into four distinct frequency bands. In the infra-gravity band, 1-10mHz, where ambient noise levels are greater than levels on the co-located buried sensor, there is a strong tidal frequency dependence indicating that water flow in the hole, past the instrument, may be exciting installation noise. The largest amplitude changes occur in the noise notch, 10-50mHz, where vertical component noise varies from the quietest levels observed world-wide [-185dB re: 1(m/s^2)^2/Hz] to levels above ?120dB re: 1(m/s^2)^2/Hz after the Balleny Islands earthquake. The micro-seism band, 50-300mHz, is characterized by three peaks. Levels of the single frequency micro-seism peak at 60mHz can increase 60dB after a large earthquake. The levels of the two double frequency micro-seism peaks, one each from distant and local sources, is much less variable (less than 20dB) and is related to sea state. The short period band (or HOLU spectrum), 300mHz to 7.5Hz, consists of a set of peaks that correspond to shear modes in the seafloor which are excited by local sea state. Above 5Hz there is a weak tidal dependent effect, primarily on the vertical component, which could be related to bottom currents washing against the re-entry cone.