Acoustic sound of bubble formation. Recorded in the laboratory.
Seep Bubble Music is the sound made by the seeps at the seabed as bubbles form. Each bubble has a tonality that is related to its size. Thus, by listening to the sound the seep bubbles make it should be possible to estimate how many bubbles are being produced, what their size is, and thus the total emission flux. However, while the methodology is well developed for the laboratory, application to the field presents numerous challenges.
Cause of Bubble Sound
When a bubble pinches off from a vent at the seabed, or from a capillary tube in the laboratory, it creates a volume oscillation in the bubble which causes a sound. This sound is quite pure and is related to the bubble size pressure, and other factors. The sound only lasts tens of milliseconds, and so sounds more like a pop. As reported in Leifer and Tang, 2007 this sound can be used to measure the size of the bubbles escaping from marine hydrocarbon seeps.
It is easy to hear the sound at home without a hydrophone. Simply take a glass and submerge it underwater in the sink. Then slowly pour the air out of the glass. the bubbles that form each have their own tone - musical pops.
A sample acoustic signal recorded in the laboratory for a bubble is shown below. The onset of the sound is extremely fast, less than a single cycle of a sound oscillation. Moreover, the sound is rapidly damped - the only source of energy is the initial formation event - and thus it lasts a very short time, sounding like a tonal pop.
|
Acoustic sound of bubble formation. Recorded in the laboratory. |
The frequency, f, of the bubble formation sound can be related to the bubble radius, r, by the Minnaert Relationship,
where gamma is the ratio of gas specific heat at constant pressure to constant volume and is either 1 - isothermal, or 5/3 - adiabatic, P is the pressure, and ro is the water density. For a r ~1-mm bubble at atmospheric pressure, f ~ 3.5 khz. the Minnaert frequency is determined by spectral analysis. For example, fourier analysis of the sound shows that the spectrum of the bubble sound is quite pure - something your ears also demonstrate.
 Spectra of bubble formation for a bubble in the laboratory calculated using standard Fast Fourier Transform, and the Burg Spectral Approximation Method. |
However, because fourier analysis, the standard approach to spectral determination, is inappropriate for such short lived sounds, other methods are required that allow the spectra to be analyzed from a short sequence. One method is the Burg method, a spectral approximator. For a system with excellent signal to noise ratio, such as achieved in the laboratory, the two spectral analysis methods agree quite well; however, there is a slight difference between the two. This difference arises in part because there is a frequency shift with time in the spectra as shown in the spectragram.
 Spectragram of bubble formation sound in the laboratory calculated using the Burg Spectral Approximation Method. Dashed red line shows peak initial frequency where the Burg Method was applied to the entire sound. |
Bubble vents were observed both optically and acoustically at the informally named La Goleta seeps (major plume, La Goleta A at 34° 22.503’ N, 119° 51.193’ W, at the seabed, 62-m deep) in the Coal Oil Point (COP) seep field. The La Goleta Seeps are characterized by scattered bubbles and small bubble plumes risiing over a vast area stretching more than a kilometer in a ENE-WSW trend and about half a kilometer wide. There are several major, intense plumes located to the southeast. The study area was located a hundred or so meters to the southeast of the main plume. See "*" in the figure below.
 Map of the Coal Oil Point hydrocaron seep field, located near the University of California, Santa Barbara in the Santa Barbara Channel. Seepage regions are shown as gray areas as determined by sonar return. The study site is denoted by the small asterisk to the lower left of the main plume of the La Goleta Seeps. Explore the seep field with the interactive seep field map, which contains information on many named seeps, including images and movies. |
The field study used the two person, manned submersible, the Delta, which was supported by the R/V Velero. Data on bubble acoustics and was collected on 21 Sept. 2005.
 The Delta submarine glodes through the water offshore of Isla Vista. Dave Slater is driving, the scientist is in the fore section. BMS was attached to the far side of the submarine. Isla Vista is in the background with Storke Tower, located on UCSB, to the far right, with the Santa Ynez Mountains in the background. |
The study area seabed was comprised of fine-grained sediment with vents separated on meter-scale distances. The vents emitted bubbles in chains or small streams. Some vents were associated with small (5-cm length scale) pockmarks in the sediment that contained tar. The vent that was analyzed is shown below to the left and emitted ~25 bubbles per second.
 Images of seabed seep vents at the La Goleta Seep. The vent shown to the right was analyzed in Leifer and Tang (2007). The video clip is available. |
To measure quantitatively the bubbles from this vent, bubbles were imaged with a bubble measurement system (BMS), which was attached to the Delta submarine. This unfortunately coupled noise very effectively from the submarine to the hydrophone. The BMS is described in Leifer and Boles 2005. More recent BMS improvements included two transparent screens to delineate the measurement volume and prevent along-axis bubble advection. Thus bubbles are within a clearly defined range of distances or size scales. Bubble blockers with downward lips prevented bubbles from rising between the camera and measurement volume or behind the screen and casting shadows ( Leifer and Tang, 2007)
  The hydrophone was attached to the BMS above a bubble blocker to prevent bubbles from hitting the transducer. To the right is an image of the BMS out the submarine window. |
Due to the noise of the submarine, the signal to noise ratio was much poorer than in the laboratory, as can be seen in the acoustic signal recorded by the hydrophone (see below). Another noteable difference is the much slower onset of the bubble sound than in the laboratory. This phenomenon has been related to bubbles in a chain ( Mansseh et al. 1998), although other factors, such as surfactants may have played a role.
 Acoustic signal of bubble formation at a marine seep vent, and subset showing a single sound. From Leifer and Tang (2007). |
As shown above, the time the bubble sound was above the noise level only lasted a few tens of milliseconds, much too short for Fourier analysis. As a result, spectragrams were calculated by the Burg Spectral Approximation Method. Two peaks are evident, at ~1700 and ~1800 Hz, separated by ~5 ms, representing two bubbles. The precise frequency of these bubbles was determined from a plot of the peak frequency with respect to time and was the frequency with the greatest spectral power (strength). In this case, there is no evidence of frequency shifting for the first bubbles, although there is a shift in the frequency of the second bubble, a few milliseconds after its formation, at ~17 ms.
 Spectragram of field bubbles using a 14-pole, 1238-order, Burg Spectral Method on 128 point data sequences with a 12 kHz sampling frequency. From Leifer and Tang (2007). |
As the video shows, many bubbles were produced nearly simultaneously, as documented in other spectragrams of other bubbles. Furthermore, where two bubbles are produced sufficently close in time and are of different size, most bubbles exhibited an acoustic coupliing and rapid frequency shift between the two dominant frequencies.
 Spectragrams of field bubbles using a 14-pole, 1238-order, Burg Spectral Method on 128 point data sequences with a 12 kHz sampling frequency. |
A dramatic frequency shift is visible in the above figure to the left where the main frequency shifts from ~1700 Hz to almost 2000 Hz in just a few milliseconds. Because the main frequency persists even as the shift is occurring, and the spectral power is much greater than was observed in the bubbles in the previous figure, a third bubble of the same size as the first may have been produced. However, signal to noise was too poor to identify bubbles of the same size that were produced simultaneously. A smaller shift is shown in the above figure to the right between 22 and 28 ms. Note, the absence of a shift in the solo bubble at 50 milliseconds. Published lab studies suggest that frequency shifts occur where bubbles are closer than about 20r where r is bubble radius. In the field, bubbles produced within 5 milliseconds, as shown above, were within 20r (~5 cm).
 Probability distributions of peak frequency (left) and bubble radius (right). From Leifer and Tang (2007). |
The peak frequencies for all the bubbles analyzed in the video were analyzed as described above, and the probability distribution of frequencies and the probability size distribution were calculated. Both were bimodal, and exhibited the same ratio between the two peaks (1.27 and 1.2). Peaks of the frequency distribution were at 1500 and 1750 Hz, while peaks in size were at 2200 and 2800 micrometers. Frequencies were about 20% lower than predicted by the Minnaert relationship. Interestingly, a 2800 micrometer bubble has about double the volume of a 2200 micrometer bubble.
Leifer I., and J. Boles, 2005. Measurement of marine hydrocarbon seep flow through fractured rock and unconsolidated sediment. Marine Petroleum Geol., 22(4), 551-568.
Leifer I., and D.J. Tang, 2007. Acoustic Signature of Marine Seep Bubbles J. Acoust. Soc. Amer. Ex. Lett., 121 (1), EL35-EL40.
Manasseh, R,, S. Yoshida, and M. Rudman, 1998. Bubble formation processes and bubble acoustic signals. In: Proc. Third Internat Conf. on Multiphase Flow, Lyon France, 8-12 June.
