Scott’s notes




Дата канвертавання25.04.2016
Памер42.75 Kb.

Scott’s notes:


Some of the conclusions are really results and/or discussion. Maybe the last section should just be “Discussion.”
Strategy:

Paper 1: Describe system and basic source level results

Paper 2: Ambient noise characterization

Paper 3: Directionality of orca calls

Future paper: Source level of orca echolocation


Title


(name, affiliation, date of upload, running title of <6wds)

Val Veirs, Colorado College, Physics Department, 14 E Cache la Poudre, Colorado Springs, Colorado, 80903

Scott Veirs, Beam Reach, 6537 16th Ave NE, Seattle, Washington, 98115
Running title:

Abstract

TEXT


Introduction


0) SR are under review as “threatened,” so it’s important to quantify their acoustic behavior and environment.

1) Calls are communication, thus we focus on them here and echolocation in a future paper.

2) Similar studies have assessed source intensities of other sounds.

3) The array is now calibrated and capable of locating sound sources which enables us to begin quantifying orca calls and echolocations, as well as the acoustic nature of their “home range.”



Methods




Results
1) Peak source level = 167dB

2) Average source level =



Discussion


1) Comparison to other odontocete source levels (any measured in aquariums? Ask Morton or others at marine parks? Or just go do it with a calibrated phone?)

2) Directionality discussion (contact Kelly Benoit-Baird re upcoming pubs or initial results?)

3) Comparison to echolocation intensity (future paper?)

4) Implications for conservation

Foraging: Theoretical level of returns and comparison to ambient intensity (future paper?)

Masking: Comparison to whale watch and commercial vessel receive levels in the orcas’ environment?



Acknowledgements

(Work supported by 32 undergraduate researchers over five years and the Colorado College Physics Department)



Appendices




Textual footnotes




References




Tables




Collected figure captions




Figures


Consider alternatives to “passby,” like encounter, event, or transit?
Source Levels of Orca (Orcinus Orca) Social Vocalizations Measured with a Shore- Based Hydrophone Array
Abstract (as submitted)
An array of four hydrophones spaced over 200 meters near the shore of San Juan Island, WA, has been used to localize the underwater vocalizations of Southern Resident orca whales. The hydrophones (ITC)have home-built pre-amps and amplifiers and signals are digitized simultaneously by two computers. The system has been calibrated with an underwater speaker (J-9). Automatic call detection is used to activate data logging on one computer and real-time localization on a second that directs a video camera to record the water's surface at the predicted source location. The system requires about one second to detect a call, locate it and redirect the video camera. This is the first report of source levels of social vocalizations of free-ranging Southern Resident orcas. To date (July 2004), 160 orca calls have been localized within 200m of the center of the array. The average acoustic intensity of these calls is 145 dB re 1microPa@1m. The range of intensities is from 137 to 157 dB re 1 microPa@1m. Complete results including intensity angular distributions with respect to the probable direction of whale travel will be presented.


I. Introduction:
Source levels1 of animals vocalizing under water are rarely reported because of the difficulty of accurately knowing the location of such sources. Additionally, questions of sound spreading and attenuation as well as hydrophone calibration must be addressed to be able to convert measured voltages to source levels. Source levels are important in the study of the meaning that may be communicated via underwater vocalizations and in assessing the impact that anthropogenic noise may have in limiting or masking such communication.
Recent reports of source levels of free-ranging marine mammals include: echolocation clicks of orcas in Johnstone Strait, B.C. range from 195-210 dB re 1Pa @ 1m for orcas echolocating on the observing hydrophone array (Au et. al. 2004), bottlenose dolphin in Moray Firth, Scotland produce whistles with mean source level of 158 dB re 1Pa @ 1m (Janik 2000), and West Indian manatee in Crystal River, Florida vocalizations average 112 dB re 1Pa @ 1m (Phillips, 2004).
Generally it is found that larger animals have louder vocalizations. Source levels for vocalizations other than echolocation clicks are reported for eight species of toothed whales in Table 7.2 of Richardson et. al. (1995). These data, which are mostly from the 1960’s through 1980’s, and the recent Moray Firth dolphin measurements are graphed versus the log of the average species body mass in Figure 1. Until further observations are made, there is considerable uncertainty regarding the accuracy and precision of these data. However, the regression line does explain 60% of the variance in the data and yields an interesting result. If we take the slope to be 20, then for these non-echolocation vocalizations, the acoustic pressure of the calls for each species is generally proportional to the average body mass, i. e. acoustic pressure (Pa) at one meter from vocalizing animal = 20,000 mass (kg).

Fig. 1. Non-echolocation vocalizations of toothed whales vs. body mass


The purpose of this study is to measure the source levels of the social vocalizations of Puget Sound’s Southern Resident orcas. This population of three pods totaling approximately 85 orcas passes by the shores of San Juan Island almost daily during the summer. From the frequency and diversity of their social vocalizations, i.e. non-echolocation click sounds, we conclude that these vocalizations play some important role in the social relations between and within the three pods. At times these sounds are immersed in loud anthropogenic noises in the same frequency range (200 Hz – 5 kHz) as the orca calls.
As we learn more about the source levels of orca vocalizations we then can estimate the ‘active space’ that they can communicate within and we can study how noise from commercial ships, whale-watch vessels, and recreational boats may affect the communication between vocalizing orca whales.

II: OrcaSound Hydrophone Array

The hydrophone array used in this study is permanently established on the west side of San Juan Island, Washington State facing Haro Strait and operates 24 hours a day all year long. Automatic signal detection is used to activate data archiving and real-time localization calculations. The Southern Resident orcas are frequent visitors and the system records vocalizations two or three times each week during the summer (May – Aug). In the course of a summer about 4 gigabytes of orca vocalizations are recorded. Figure 2 shows the locations of the four array hydrophones and shows a typical location calculation, signifying an orca vocalizing at a depth of 7 meters.



A: Hydrophones

The OrcaSound hydrophone array consists of one ITC-6050C hydrophone and three ITC-4066 hydrophones, both made by International Transducer Corporation (www.itc-transducers.com). The 6050C hydrophone is equipped with an internal preamp while the 4066 hydrophones use a custom built preamp designed to eliminate noise. These custom preamps, built by Colorado College students, use an AD524 chip to produce a gain of 100. The hydrophones are sensitive to frequencies ranging from 100 Hz-10 kHz, covering the range of most boat noise and orca vocalizations. The cables from the four hydrophones come through the intertidal inside cast-iron cable protectors which prevents their destruction by storms. On shore, another stage of amplification is applied and then when the signals get to the data gathering and analysis computers, isolation amplifiers are used to eliminate the common-mode electrical noise that is generated with the long cable runs. The frequency range of these hydrophones and amplifiers is from 100 Hz to 10,000 Hz. The signals are digitized at 22,050 Hz. The sensitivity of each of these hydrophone and amplifier systems is approximately –105 dB re 1V/Pa. The calibration of these hydrophones is discussed later.


A computer with an A/D board maintains a ringbuffer of the previous minute of data from each hydrophones at all times. A second computer examines two of the hydrophone channels and listens for “interesting” sounds. When triggering criteria are met, a signal is sent to the first computer and that machine saves the recent data and continues to record as long as triggering is maintained. Once a trigger has been initiated, localizations are performed.

B. Signal Processing


The triggering algorithm that OrcaSound uses is quite straightforward. Two different running averages of background sounds are calculated for each of the two channels used for triggering. A 1 minute averaging time gives the “background” sound level and a 1 second averaging time gives the “signal” sound level. The ratio of these two is the signal to noise ratio. If this S/N ratio rises above a threshold and then descends below a lower threshold (Schmidt trigger) in a time interval longer than 0.5 sec and less than 2 sec and this happens simultaneously on two hydrophones, then the system triggers. This algorithm reliably triggers on the orca calls of interest independent of the quite variable background noise levels at this site. It also triggers on other sounds with this envelope profile, most frequently the heavy breathing sound of a nearby harbor seal and occasionally the SONAR pings of U. S. Navy ships.



FIG. 2. Hydrophone locations off west side of San Juan Island, WA. The grid scale is in meters and the clock at the upper left shows the time of the passby.

In order to compute the location of an underwater source, it is necessary to measure the small time and spatial differences between when signals get to one hydrophone compared to each of the other hydrophones. Figure 3 diagrams these extra distances that are measurable with this hydrophone array.

Fig. 3. Cartoon showing the extra distances that sound travels. These are the measurables in this experiment.


The amplitude envelopes on two hydrophones of two typical orca vocalizations are shown in Figure 4a. Figure 4b shows a 2 second interval centered on the second call shown. The maximum signal ranges from 113 dbRe 1 Pa on hydrophone 0 to 117 dB re 1 Pa on hydrophone 2.
The signals shown in Figure 4b show how subtle these differences are. The crosspower spectrum of pairs of signals is used to compute the time difference between the arrival of a call at one hydrophone and another. Figure 4c shows the crosspower spectra for pairs of hydrophones (0,1), (0, 2), and (0,3). The maximum of the crosspower spectrum is the best estimate of the time lag between the two arrivals. In the spectra shown, these time differences range from 20 ms to 49 ms with the signal arriving latest at hydrophone 0. Using the speed of sound in water, typically 1480 m/s, these time differences can be converted into the extra distances in space that a signal has traveled from its source to hydrophone 0 compared to travel to each of the other hydrophones.



Fig. 4a. Amplitude vs time in seconds for two hydrophones.


Fig. 4b. Time series for four hydrophones and r.m.s. amplitude background levels and maxima for each. Horizontal axis is samples at 22,050 samples/sec.


Fig. 4c. Crosspower spectra for three pairs of signals showing the offset between the pairs of signals in samples, milliseconds, and distance. The Q is a measure of the quality of the crosspower spectrum.


The Q value reported in Fig. 4c is the height divided by the width (full width half maximum) of each crosspower spectrum. This quality value is used to select ‘good’ data where precise localizations can be calculated.

Source localization is perform using “matched field processing”. Matched field refers to a nonlinear minimization of an error function composed of a comparison between the observed time intervals and those predicted via straight line ray sound wave propagation from a hypothetical underwater location to each of the hydrophones in the array:



Using a multidimensional downhill simplex algorithm (Press, 2002)2, this hypothetical location is moved around until the root mean square error between the measured time differences and those predicted for the hypothetical location is minimized. It typically takes a few hundred error function evaluations to find the minimum and this takes less than ~1/2 sec on our Pentium III computer.

C. System Calibration

The OrcaSound hydrophone array was calibrated and the locations of the hydrophones was determined using an underwater sound projector and a GPS receiver. The underwater sound projector was a J-9 and was calibrated using an F-42b hydrophone3. The source level of the J-9 was operated at 140 dB re 1 Pa @ 1 m over frequencies from 500 Hz to 3 kHz at many locations and depths in the region of the hydrophone array. A CTD was also used to measure the speed of sound as a function of depth in the region of the array. One-half second chirps with frequency rising from 500 Hz to 3 kHz were sent into the water from measured depths. GPS waypoints were taken simultaneously with each chirp. These data are used to (1) determine the locations of the hydrophones, (2) determine the calibration of each of the four hydrophone-preamp-cable-amplifier systems, and (3) determine the way that sound spreads out as it proceeds from source to hydrophone in this shallow water environment.


Approximate hydrophone locations were measured when the hydrophones were initially lowered from a small boat. GPS readings were taken and the depth of each hydrophone was measured with the rope used to lower each weighted stand with attached hydrophone. However, these observations were made several years ago and visits by SCUBA have shown us that over the course of a year, some hydrophones have moved. (Actually, several have gone missing, with the heavy marine cable apparently cut.)
To improve on these approximate locations and to check on the locations every few months, about 100 chirps described above are localized. An error function is composed of these predicted locations compared to the GPS and depth locations of the underwater projector at the moment of each chirp. A non-linear minimization of this error function is carried out by moving the hydrophones around in 3 dimensional space until a minimum is found. This is also done with a multidimensional downhill simplex algorithm. Several hours of Pentium III computing is needed for a satisfactory solution. Table I shows the accuracy of localizing the underwater speaker for 50 chirps spread out over a 200 m (east-west) by 400 m area (north-south). There is a systematic error in underdetermining locations about 250m south of the array which leads to the asymmetric errors shown. The overall error uncertainty of a location is about 10 m in both the east-west and up-down directions and nearly twice this in the north-south direction.





Mean Error (m)

Standard Deviation (m)

East-West distance difference

1

11

North-South distance difference

8

27

Vertical distance difference

0

8

Range difference

8

23

Table I: Known Distance Minus Predicted Distance: Means and Standard Deviations

Figure 5 shows the predicted and GPS locations for these 50 chirps.

Figure 5: The black circles are GPS locations of chirps and the blue circles are locations determined with the cross power spectral determination of travel time differences and matched field processing using the computed hydrophone locations. The red circles are the hydrophone locations and the scale is in meters. The shoreline trends northwest.

(Note: I need a better picture and I need to ‘fix’ the systematic error.)
These same chirps are used to calibrate the hydrophones and to determine what spreading model is appropriate in this near shore location. The J-9 projector was operated with a source level of 140 dB re 1 Pa @ 1 m as determined with a calibrated F-42b Navy hydrophone. Careful observations of the strength of the chirps received from the J-9 were made, as the J-9 was rotated 360 degrees in a horizontal plane. No variation in intensity was observed as a function of orientation. Hence, we can use this known source level to both calibrate the hydrophone/preamp/amplifier/digitizer systems and to determine the appropriate spreading model.
Figure 6 shows the result of this calibration for hydrophone number 1. This graph shows then received level of the hydrophone versus the log (base 10) of the distance from the hydrophone to the J-9 projector for each of the chirps. The straight line is the best fit line and shows that the source level of the J-9 is 140 dB, the intercept, and the appropriate spreading model has a slope of –9.9 dB per decade in distance, which is cylindrical spreading. The sensitivity of the hydrophone/amplifiers system for this channel turns out to be –128 dB re 1 Pa/V. Similar results were obtained for each of the channels.

Figure 6. Hydrophone 0 Calibration and Spreading Law
The source levels determined for these chirps from all of the hydrophones are histogrammed in Fig. 7. The red dots are the number of observations (200 total) of the source level of the J-9 projector. The green curve is a gaussian distribution that fits these data extremely well. The mean source level of the J-9 is found to be 140 dB and the standard deviation of the distribution about this mean is 2 dB.

Figure 7. J-9 Projector Computed Source Levels from 50 Chirps.

IV: Conclusions

Recent reports of source levels of free-ranging marine mammals include: echolocation clicks of orcas in Johnstone Strait, B.C. range from 195-210 dB re 1Pa @ 1m for orcas echolocating on the observing hydrophone array (Au et. al. 2004), bottlenose dolphin in Moray Firth, Scotland produce whistles with mean source level of 158 dB re 1Pa @ 1m (Janik 2000), and West Indian manatee in Crystal River, Florida vocalizations average 112 dB re 1Pa @ 1m (Phillips, 2004).


Generally it is found that larger animals have louder vocalizations. Source levels for vocalizations other than echolocation clicks are reported for eight species of toothed whales in Table 7.2 of Richardson et. al. (1995). These data, which are mostly from the 1960’s through 1980’s, and the recent Moray Firth dolphin measurements are graphed versus the log of the average species body mass in Figure 9. Until further observations are made, there is considerable uncertainty regarding the accuracy and precision of these data. However, a regression line does explain 60% of the variance in the data and yields an interesting result. If we take the slope to be 20, then for these non-echolocation vocalizations, the acoustic pressure of the calls for each species is generally proportional to the average body mass, i. e. acoustic pressure (Pa) at one meter from vocalizing animal = 20 mass (metric tons)4.


Fig. 9. Non-echolocation vocalizations of toothed whales vs. body mass
The results of this study are shown in Figure 9. The arrows signify the two standard deviation ranges of Southern Resident orca social vocalizations.
The 445 orca vocalizations localized in this study come from 24 passbys of the fixed hydrophone array over a three-month period. Figure 10 shows the mean source level of orca vocalizations for each of the passbys graphed versus the number of calls per minute during each passby. The errors shown are the standard error of the mean source level for each of the passbys. It appears that passbys with a low call rate divide into two groups with significantly different source levels and source levels may tend to rise as the call rate increases.

Fig. 10. Southern resident source levels versus call rate for 24 passbys


Only two passbys occurred after sundown. One at midnight on Sept. 1, 2004 and one at 3 am on July 10, 2004. The average source level for both of these after-dark passbys is significantly lower than for the daylight passbys.
Figure 11 shows the average call source level for each passby plotted versus the average background noise level during that passby. The average volume of orca vocalizations tends to increase with increasing background noise. The two night-time passbys are not far off the regression line. Hence, it appears that on these two night-time occasions, the orcas vocalized more softly even though the background noise was not exceptionally low.

Fig. 11. Average source level of passbys versus average background noise level


The loudest source level recorded was 167.8 dB re 1Pa @ 1m recorded at 9 am on July 6, 2004. This passby contained the largest number of calls, 66, and had the second highest average call source level, 151 dB re 1Pa @ 1m, and intermediate background sound levels, 107 dB re 1Pa. Figure 12 shows the frequency distribution (sonogram) for this call. This upsweep in frequency is classified as S-19 in the Ford – Osborne (Ford 1987, Osborne yyy) southern resident call category scheme.

Fig. 12 Sonogram of loudest call – 167.8 dB re 1Pa @ 1m


Figure 13 shows the time series (recorded voltage vs time) for three orca calls. The orca call that had the highest calculated source level is the second of these three.



References:
Au, W. W. L., Ford, J. K. B., Horne, J. K., Newmann Allman, K.A., (2004) “Echolocation signals of free-ranging killer whales (Orcinus orca) and modeling of foraging for Chinook salmon (Oncorhynchus tshawytscha)”, J. Acoust. Soc. Am. 115, 901-909.
Janik, V. M., (2000) “Source levels and the estimated active space of bottlenose dolphin (Tursiops truncates) whistles in the Moray Firth, Scotland”, J. Comp. Physiol. A 186, 673-680.
Phillips, R., Niezrecki, C., and Beusse, D. O., (2004), “Determination of West Indian manatee vocalization levels and rate”, J. Acoust. Soc. Am. 115, 422-428.
Press,W.H., Teukolsky, S. A., Vetterling, W. T. and Flannery, B. P., (2002) Numerical Recipes in C++, Cambridge University Press, Cambridge, UK.
Richardson, W. J., Greene, C. R., Malme, C. I., Thomson, (1995) D. H. Marine Mammals and Noise, Academic Press, New York.



1 Source level refers to the intensity of sound at a specified reference distance from a source. The usual distance used is 1 meter and levels are reported in decibels relative to a reference pressure of 1 microPascal (Pa) and therefore, source levels are expressed as dB re 1Pa @ 1m.

2 The simplex algorithm was re-coded in Visual Basic in the style of the Press et. al.’s C++ code

3 The J-9 projector and calibrated F-42b hydrophone were rented from the Naval Undersea Warfare Center in Newport, RI.

4 The orca source levels reported here are not included in the regression calculation.





База данных защищена авторским правом ©shkola.of.by 2016
звярнуцца да адміністрацыі

    Галоўная старонка