Singing icebergs
Harmonic tremors and their analogies to volcanic tremor
From July until November 2000, several episodes of sustained seismic signals were recorded at the seismological network at Neumayer Base near the continental margin of Dronning Maud Land/Antarctica. The signals of up to 16 hours duration were recorded at a distance of up to 820 km from the source. Their spectra show narrow peaks with fundamental frequencies ranging from 0.5 - 6 Hz, up to 30 integer harmonic overtones, and frequency gliding, resembling volcanic tremor. However, frequency-wavenumber analysis from recordings of the small aperture array VNA2 suggested a moving source, which was recognised to be iceberg B-09A travelling along the coast of East Antarctica.
Calculation of reduced displacement to estimate the strength of the tremor signals obtained values comparable to strong volcanic tremor observed for example at Kilauea and Mount St. Helens.
Studying the source processes of iceberg tremor signals may give clues on the still poorly understood tremor-generating mechanisms inside volcanoes. The most probable tremor source are fluid-flow induced vibrations inside the iceberg's tunnel/crevasse-systems. Cracks and crevasses in glacial ice may well have such geometry as do dikes in a volcano. However, whereas viscous magma has to be forced with enormous pressure through the cracks to reach the critical velocity, the viscosity of water and hence the pressure needed to initiate tremor is at least three orders of magnitude lower. It is supposed that the flow speed of the Antarctic Coastal Current is sufficient to build up the driving pressure for tremor if the iceberg’s free drift is hindered for example by collision with the seafloor as illustrated by the tremor episode of July 22. As seen from a high resolution satellite image (Fig. 1B), B-09A is indeed highly crevassed, and oscillations of the iceberg in free water or changing approach directions of flow following collision events may be responsible for the various stages of tremor evolution.
In the first half of 2006, again several iceberg trmor episodes could be recorded which originated from large icebergs B-15D, C-08, D-18 and D-19.
Publication
Müller, C., Schlindwein V., Eckstaller A. & Miller, H. 2006. Singing icebergs, Science, 310, 1299.
Figures
Figure 5
A) Nearly 2 hours of iceberg tremor recording from VNA2 at 19 April 2006 and corresponding spectrogram. B) Result of array fk-analysis of tremor event which reveals backazimuth corresponding roughly to positions of icebergs C-08 and B-15D (C, ENVISAT-ASAR satellite image).
Figure 4
VNA2 array beam of the two the 22 July 2000 tremor preceding earthquakes and corresponding spectrogram. The first appears like a normal tectonic event with clear P- (A) and S-wave (B) onsets, broad-band frequency content, and rapidly decaying coda. The second event (C) however, following nearly three minutes later is characterised by lacking S-waves (D) and slowly decaying coda with spectral peaks at 4, 8, and 12 Hz. Here, also slight frequency gliding occurs. The slowly decaying coda and the spectral behaviour are comparable to volcanic 'tornillo'-type events.
Figure 3
Comparison of the seismograms and power spectral densities (PSD) for stations SNAA, VNA1, VNA2, VNA2 array beam, and VNA3 for tremor episodes on 22 July 2000 (A) and 16 July 2000 (B). Shown are 40 s time windows of vertical component records and corresponding power spectral densities. A) The PSD is dominated by nearly 30 sharp spectral peaks at 0.96 Hz, 1.44 Hz, 1.92 Hz, ... at intervals of 0.48 Hz. This spectral behaviour suggests a fundamental frequency at 0.48 Hz which is masked by long-period noise. Note the alternating high and low spectral amplitudes known as period doubling. This can best be seen at higher frequencies of the VNA2 array beam. B) Same for the tremor of 16 July, showing a fundamental frequency of 1.65 Hz and again, exactly integer harmonic overtones. The peaks occur at exactly the same frequencies for all stations, indicating a source origin of the signals rather than path or recording site effects.
Figure 2
Tremor seismograms and corresponding spectrograms of the events recorded on 22 July 2000 (A) and 16 July 2000 (B). Array beams are formed using backazimuth and slowness values estimated from array-fk-analysis. The seismograms are highpass filtered with a corner frequency of 0.5 Hz to suppress long-period noise. A) The tremor on 22 July 2000 was the most spectacular signal with long duration, high amplitudes, and a variety of spectral features. The figure shows a record of 18.3 hours duration. Irregular tremor is already present before the strong increase of amplitudes initiated by two 'earthquakes' (A) followed by an about two hours lasting sequence of seismic events with strongly varying frequency contents (A-B). After about one hour of quiescence (B-C) the tremor starts with discrete frequencies which are slightly decreasing and broadening with time and increasing amplitudes (C-D). At time D, a transition from harmonic to chaotic tremor occurs with still increasing amplitudes lasting for about 3.5 hours. From E to F, several transitions between chaotic and harmonic tremor can be observed, also showing period doubling phenomena. Harmonic tremor abruptly ends at F. B) The tremor on 16 July 2000 started with chaotic spectral behaviour, transition to sharply peaked harmonic tremor with strongly decreasing and again increasing frequencies (A-B). B-C again shows chaotic behaviour and a very abrupt ending at C.
Figure 1
The stations of the Neumayer Base seismological network, the iceberg B-09A, its track and tremor localisation. A) Satellite image (Advanced Very High Resolution Radiometer (AVHRR), thermal image) of B-09A at 5 August 2000 offshore western Dronning Maud Land and the sites of the four seismological stations of the Neumayer seismological network. B) Synthetic aperture radar satellite image (RADARSAT-1 satellite) of B-09A in 1997. The dimensions of the iceberg are approximately 30 to 50 km. Compared to low-resolution satellite images during tremor episodes in August 2000, the iceberg has lost the dashed part to the right. Note the extended crevasses seen on the surface of the iceberg. The inset shows the travel path of B-09A from beginning of 1992 until February 2003 in the westward Antarctic Coastal Current. B-09 calved off the eastern Ross Ice Shelf in late 1987, undergoing a long history with splitting into B-09A and B-09B, and B-09A being stranded twice for several years until passing Dronning Maud Land. C) Western Dronning Maud Land showing daily averaged positions of B-09A from QuikSCAT satellite radar backscatter images. The size of the circles approximate to the dimension of the iceberg. The blue line marks 350 m water depth corresponding to the assumed draught of the iceberg. The positions of the iceberg at the times of harmonic tremor signals are highlighted. The inset shows the geometry of the small aperture array VNA2. The backazimuths of the tremor signals were estimated by fk-analysis of the array recordings and are shown as red lines. The tremors’ backazimuths clearly follow the track of the iceberg. Slight directional mislocations are due to unknown crustal heterogeneities. At 17 and 18 July and between 28 July until 26 August 2000 the array was not operational. The two stars indicate the epicentres of the two distinct events preceding the strong tremor at 22 July 2000 as derived from P- and S-wave onsets at all network stations.