Summary. Methodology. Selected field examples of the system included. A description of the system processing flow is outlined in Figure 2.

Transcription

1 Halvor Groenaas*, Svein Arne Frivik, Aslaug Melbø, Morten Svendsen, WesternGeco Summary In this paper, we describe a novel method for passive acoustic monitoring of marine mammals using an existing streamer spread of point receivers, consisting of seismic and positioning hydrophones. Together, these hydrophone sets have a significant overlap with the frequency band containing the calls of the mammals we are trying to detect. The exceptional spatial extent of the spread along with the large number of sensors yields superior localization capabilities Selected field examples of the system in action are included. proposed. Such systems must have a large enough aperture to obtain high resolution and a high number of receivers to achieve sufficient signal-to-noisee ratio. Developing, deploying, and operating such systems reliably, therefore, poses significant technical, operational, and logistical challenges. This is especially true when these systems must be towed in parallel with the seismic insea equipment. By having the system integrated into the streamer spread, the operational complications are avoided and accurate localization performance is consistently achieved. Methodology A description of the system processing flow is outlined in Figure 2. Figure 2 Flowchart of the passive marine mammal monitoring system. Figure 1. Example from the field: The mammal monitoring system tracking a sound source. The grey stars indicates detections over a certain period of time. Introduction Marine mammals such as whales, dolphins, and porpoises use acoustic signals for navigation, communication, and hunting. There is an ongoing discussion how the energy radiated from seismic sources might impact the communication and behavior of marine mammals. The seismic industry follows IAGC guidelines and local regulations that normally include soft start-up procedures, exclusion zones around the seismic source, and presence of marine mammal observers (MMO) onboard seismic vessels. To aid these observers during periods of limited visibility, passive acoustic monitoring systems have been Data Acquisition Continuous data are recorded from the full seismic spread using point receiver seismic hydrophones at m intervals along the streamer and covering the lower part of the frequency spectrum, while point receiver positioning hydrophones at approximately 60m spacing cover the high- detection frequency part of the spectrum. Signal-to-noise enhancement and signal The large aperture and sensor count of the array is used to improve the signal-to-noise ratio of the data, detect any coherent signals, and extract observations that serve as input in the localization algorithm. Seismic data are characterized by high noise levels at lower frequencies due to streamer vibrations, vessel noise, wave action, and other noise sources. The seismic source itself also significantly decreases the signal-to-noise ratio levels at the receivers. To achieve sufficient signal-to-noise ratio, it is necessary to take advantage of the large aperture and number of sensors of the seismic streamer. Beamforming, implemented as a frequency-wavenumber (f-k) estimation is used for this purpose. SEG San Antonio 2011 Annual Meeting 62

2 Figure 3 Low frequency data processing. Top: Spectrogram of Pygmy whale call with two main harmonics at 23 Hz and 70Hz. Bottom: f-kk spectrum from simulation of a subsection receiving the call of a pygmy whale. The peaks in the f-k spectrum match the frequencies in the call. array gain. Instead, cross-correlation of hydrophone traces is used to enhance the signal-to-noise ratio. The detection part consists of picking the peaks of the correlograms and assigning them to different arrival paths. The extracted observations are the time differences of arrival between the hydrophones. Figure 4, illustrates the high-frequency data processing for detecting marine mammals. The recordings are converted to time - frequency spectra, by means of the short-time discrete Fourier transform (figure 4, left). These spectrograms are then correlated to detect similar signals on the two hydrophones. The peak locations gives the relative travel times used for inversion in the localization step (figure 4, right). Mammal localization The localization algorithms invert the observations into a position for the whale. This is done by fitting the measured observations to those predicted by a model and an assumed predefined whale location. For the low-frequency system, this consists of fitting the observed wavenumber from the subsection to a propagation model that allows both directand bottom-reflected arrivals. This is posed as a leastwhere we only would squares problem. In the special case have two subsections, this is a triangulation. For the high-frequency system, we similarly take the extracted time difference of arrival measurements and fit them to a propagation model. When processing low-frequency data, the streamers are divided into a number of subsections. An f-k spectrum is generated for each subsection. Coherent signals present on several hydrophones will appear as peaks in the f-k spectrum with the wavenumber corresponding to the incidence angles. The detection consists of picking these peaks from all subsections. Figure 3, top shows the time-frequency plot of a pygmy whale call. By computing the f-k spectrum of such a subsection 'array', we get the F-K spectrumm shown on the right. The peaks have frequencies corresponding to the tones of the call used for modeling, 23 Hz and 70 Hz, and represent the observations used as input to the localization algorithm. The positioning hydrophones are spaced too far apart for beamforming to be an alternative for the high-frequency system. However, the higher-frequency bands are less affected by noise, and we are not as dependent on large Figure 4 High-frequency data processing. Two simulated traces with humpback whale calls (left) are transformed to spectrograms (middle) using short-term Fourier transforms. The spectrograms are correlated to produces a correlogram (right), from which time differences of arrivals can be estimated. To verify the system performance, a framework was made to test calibrated whale signatures mixed into real hydrophone data using an acoustic propagation model. This procedure provided very realistic synthetic data that were used to develop and test the system. SEG San Antonio 2011 Annual Meeting 63

3 A 3D location is defined for both the whale and the sets of receivers to be used for localization. Then, the wave equation is solved for the source-receiver geometry using realistic geo-acoustic parameters. Finally, the whale signature is merged into the hydrophone traces to create a data set that is used to test the system. These simulations verified that there is a very good agreement between estimated and actual whale locations. Figure 5 shows a localization example of a whale signal that has been embedded into real seismic data. The modeling was done using a full solution of the wave equation. Subsequently, the data were processed using the low-frequency algorithm described earlier.. The accuracy of the algorithm is typically better than 300m. The system also comprises QC tools which help users confirm that the tracked signal originates from a whale and not a spurious sound source in the ocean. Quality control The system will try to locate any point sound source in the sea surrounding the seismic spread. Mammals are not alone in emitting acoustic energy into the water: The tow vessel, shipping, construction work and offshore activity can all be sources of unwanted detections. To enable operators to discriminate between marine mammals and other sound sources, the FK-plots and spectrograms are visualized for close inspection. Figure 6 Show an ensemble of spectrograms as seen on the operator terminal. It contains all the spectrograms from the high-frequency array for a single shot. This overview allows the operator to quickly decide if an actual whale was detected. Similar plots exist for the low-frequency system,, where the fk-plots of all the groups are shown together. Figure 7 similarly shows a close view of single spectrogram calculated from the signal recorded in the field on a single positioning hydrophone. The plot shows what a typical whale call might look like. Similar events will be present in all spectrograms if the detection is an actual whale call. Other sound sources, like vessel noise, has quite different characteristics to this and can hence be easily distinguished from the whale detections. Figure 8 shows an example F-K K spectrum from a field trial of the system,, calculated from the signals recorded from the seismic hydrophones in a single group. Figure 5. Simulation of a whale localization using the seismic Real-time hydrophones. visualization The red square is the true location and the blue cross is the estimated location. The error ellipse is blue. The horizontal lines indicate the seismic streamers. The location and extension of the subsections used for beamforming are indicated by red lines. The detection array consists of subsections on the two outer streamers together with a single center streamer. The full seismic array consists of ten streamers, and is not shown here. After the marine mammal location has been determined, the results are visualized in real time, using displays that show the position of the vessel, the source, and the streamers. By plotting the location of the mammals in this display, both the marine mammal observer and the maritime and seismic crew have access to the same information. Figure 6. Snap shot from QC panel of operator showing the spectrograms of all the hydrophones of the HF array. To verify that an external sound source was actually detected, the operator will typically look for shapes that are similar across the arrays. To conclude that the sound source was a marine mammal, the shapes of the detection as visualized in the spectrogram must be interpreted. SEG San Antonio 2011 Annual Meeting 64

4 The detections can be seen as a series of pink stars. Figure 7. Single Spectrogram of a whistle sound recorded by the system in the Indian Ocean usingg seismic streamer positioning hydrophones. This type of rich time-frequency structure is typical of humpback whale calls. The signal-tonoise ratio is adequate to get a proper detection. Figure 9. The monitoring system is tracking the airgun signal coming from an undershooting vessel (not displayed) running in parallel with the main tow vessel. The detections are indicated with pink stars on the display. The beige circle indicate the exclusion zone around the source Conclusion We presented a new method for performing passive marine mammal monitoring using seismic and positioning hydrophones within a seismic spread. The system is currently deployed in the field and will aid marine mammal observers, operators, and clients in executing seismic operations that minimize impact on marine mammals. Figure 8. Frequency-wavenumber spectrum from a single hydrophone group. The spectrum has been filtered with an LC filter at 30Hz and all energy outside the signal cone has similarly been removed.. This is indicated with red lines. The system is fully integrated into the seismic spread, and involves no additional deployment of special equipment. It poses minimal operational and logistical challenges to operations. The key advantage is the huge aperture and large number of closely spaced point receiver hydrophones, which enable accurate localization. The capabilities of the system has been verified in the field. Field example To verify the system in the field, we set the system up to track various sound sources for which we had knowledge of the real position. One such example is shown below, in Figure 9. Here, the high-frequency system is tracking the airgun signal coming from a vessel performing an undershoot. The undershoot vessel (not shown in the display) is steaming in parallel with the main tow vessel. SEG San Antonio 2011 Annual Meeting 65

5 EDITED REFERENCES Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2011 SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for each paper will achieve a high degree of linking to cited sources that appear on the Web. REFERENCES Richardson, J., C. R. Greene Jr., C. I. Malme, D. H. Thomson, 1995, Marine mammals and noise: Academic Press. Perrin, W. F., B. Wursig, and J. G. M. Thewissen, 2008, Encyclopedia of marine mammals: Academic Press. Jensen, F., W. A. Kuperman, M. B. Porter, and H. Schmidt, 1997, Computational ocean acoustics: AIP. SEG San Antonio 2011 Annual Meeting 66