Ambient noise in the Bedford Basin

Transcription

1 Defence Research and Development Canada Recherche et développement pour la défense Canada Ambient noise in the Bedford Basin Kabir Chattopadhyay Sean P. Pecknold DRDC Atlantic Research Centre Defence Research and Development Canada Scientific Report DRDC-RDDC-2016-R112 August 2016

2

3 Ambient noise in the Bedford Basin Kabir Chattopadhyay Sean P. Pecknold DRDC Atlantic Research Centre Defence Research and Development Canada Scientific Report DRDC-RDDC-2016-R112 August 2016

4 Her Majesty the Queen in Right of Canada (Department of National Defence), 2016 Sa Majesté la Reine en droit du Canada (Ministère de la Défense nationale), 2016

5 Abstract To facilitate the development of better underwater communications technology, a sea trial was carried out in the Bedford Basin (NS, Canada) in the summer of Modems were used to send messages and coincident ambient noise measurements were recorded. The ambient noise data were analyzed in order to verify the signal to noise ratios of the equipment and to investigate the characteristics of the Bedford Basin noise during this time. Specifically, the ambient noise power spectra were compared to two existing ambient noise models. The model-data agreement for each comparison varies significantly by time of day and frequency band. In general, the Merklinger-Stockhausen model was more accurate than an extension of the wind-noise part of the APL-UW model. However, there were fairly large discrepancies between the model and data even on an averaged basis. Significance for defence and security The level of underwater ambient noise is one of the most important features constraining underwater communications capabilities. The Bedford Basin is one of the primary test areas for underwater communications for the Force ASW (Anti-submarine warfare) project, therefore knowledge of the ambient noise there is necessary for testing equipment for Force ASW. DRDC-RDDC-2016-R112 i

6 Résumé Pour faciliter l élaboration d une meilleure technologie des communications sous-marines, on a procédé à un essai en mer dans le bassin de Bedford (N.-É, Canada) pendant l été Des modems ont été utilisés pour envoyer des messages et on a consigné les mesures du bruit ambiant coïncident. Les données sur le bruit ambiant ont été analysées afin de vérifier les rapports signal sur bruit de l équipement et pour étudier les caractéristiques du bruit dans le bassin de Bedford à l époque. Plus particulièrement, on a comparé les spectres de puissance du bruit ambiant à deux modèles de bruit ambiant existants. La concordance entre les valeurs des modèles et celles des données pour chaque comparaison varie considérablement selon le moment de la journée et la bande de fréquence. En général, le modèle Merklinger-Stockhausen était plus précis que le prolongement du volet sur le bruit du vent du modèle APL-UW. Toutefois, il y avait des écarts assez importants entre le modèle et les données, même de façon générale. Importance pour la défense et la sécurité Le niveau de bruit ambiant sous-marin est l une des caractéristiques les plus importantes qui restreignent les capacités des communications sous-marines. Le bassin de Bedford est une des principales zones d essai sur les capacités de communications sous marines pour le projet de guerre anti-sous-marine (GASM) de la force ; par conséquent, des connaissances sur le bruit ambiant à cet endroit sont nécessaires pour la mise à l essai de l équipement dans le cadre du projet GASM de la force. ii DRDC-RDDC-2016-R112

7 Table of contents Abstract i Significance for defence and security i Résumé ii Importance pour la défense et la sécurité Table of contents List of figures ii iii v List of tables viii 1 Background Experiment Experimental setup Experimental procedure Data analysis Environmental data Underwater environment Weather conditions Ambient noise data Power spectral levels Ambient noise time dependence Noise statistics Noise modelling Ambient noise models Model-data comparisons Conclusions DRDC-RDDC-2016-R112 iii

8 References Annex A: Ambient noise spectrograms iv DRDC-RDDC-2016-R112

9 List of figures Figure 1: Figure 2: Figure 3: Figure 4: Figure 5: Bathymetry of the Bedford Basin and locations of four modem nodes (3, 5, 6, 7) and Whalesong recorder Teledyne Benthos modem (top) and SM2M (Whalesong) recorder (bottom) Temperature profiles in the Bedford Basin during the period 24 July 13 August The colour of the profile changes from red at the beginning of the trial through the rainbow to purple at the end of the trial Salinity profiles in the Bedford Basin during the period 24 July 13 August The colour of the profile changes from red at the beginning of the trial through the rainbow to purple at the end of the trial Sound speed profiles in the Bedford Basin during the period 24 July 13 August The colour of the profile changes from red at the beginning of the trial through the rainbow to purple at the end of the trial Figure 6: Wind speeds during the period 24 July 13 August Figure 7: Relative precipitation levels during the period 24 July 13 August Figure 8: Figure 9: Figure 10: Figure 11: Top: Mean power spectrum and percentiles for Channel 1 over the entire experiment. Bottom: Mean and percentiles of third-octave sound pressure level (black line) Top: Mean power spectrum and percentiles for Channel 2 over the entire experiment. Bottom: Mean and percentiles of third-octave sound pressure level (black line) Channel noise level difference. The black line shows the theoretical effect of the two-pole Butterworth filters at 3 Hz and 180 Hz, and the red line shows the average spectrum of Channel 2 minus Channel 1. An adjustment of 1.1 db has been added to the difference Ambient noise spectrogram (Channel 2) and wind speed (black line) on 25 July The logarithmic frequency is given on the left y-axis, and the wind speed in m/s on the right y-axis. The x-axis shows the time of day in hours DRDC-RDDC-2016-R112 v

10 Figure 12: Figure 13: Third-octave SPL for the band 4064 Hz to 5120 Hz vs. measured wind speed. Green line shows the least-squares fit between the logarithm of the wind speed and the SPL Pearson correlation coefficient R 2 vs. frequency for third-octave SPL correlation with the logarithm of wind speed Figure 14: Histogram of SPL for third-octave band 806 Hz to 1016 Hz Figure 15: Skewness and excess kurtosis vs. frequency for third-octave bands Figure 16: Figure 17: Figure 18: Figure 19: Figure 20: Figure 21: Figure 22: Figure 23: Typical ambient noise spectra from turbulence, molecular agitation (thermal), shipping, and wind Example comparison of both models using a NUSC shipping level of 4 and a wind speed of 10 m/s. For the APL-UW model, no temperature difference between sea surface and air is used Median ambient noise levels in Hz band and corresponding NUSC shipping levels Ambient noise data minus Merklinger-Stockhausen model results with respect to frequency for period of the experiment Ambient noise data minus APL-UW model results with respect to frequency for period of the experiment Difference spectrogram between the data and Merklinger-Stockhausen model vs. time of day Difference spectrogram between the data and APL-UW model vs. time of day Third-octave mean of data minus model for Merklinger-Stockhausen model and APL-UW model. Error bars show one standard deviation. 27 Figure A.1: Ambient noise spectrogram and wind speed for 24 July Figure A.2: Ambient noise spectrogram and wind speed for 25 July Figure A.3: Ambient noise spectrogram and wind speed for 26 July Figure A.4: Ambient noise spectrogram and wind speed for 27 July Figure A.5: Ambient noise spectrogram and wind speed for 28 July Figure A.6: Ambient noise spectrogram and wind speed for 29 July vi DRDC-RDDC-2016-R112

11 Figure A.7: Ambient noise spectrogram and wind speed for 30 July Figure A.8: Ambient noise spectrogram and wind speed for 31 July Figure A.9: Ambient noise spectrogram and wind speed for 1 August Figure A.10:Ambient noise spectrogram and wind speed for 2 August Figure A.11:Ambient noise spectrogram and wind speed for 3 August Figure A.12:Ambient noise spectrogram and wind speed for 4 August Figure A.13:Ambient noise spectrogram and wind speed for 5 August Figure A.14:Ambient noise spectrogram and wind speed for 6 August Figure A.15:Ambient noise spectrogram and wind speed for 7 August Figure A.16:Ambient noise spectrogram and wind speed for 8 August Figure A.17:Ambient noise spectrogram and wind speed for 9 August Figure A.18:Ambient noise spectrogram and wind speed for 10 August Figure A.19:Ambient noise spectrogram and wind speed for 11 August Figure A.20:Ambient noise spectrogram and wind speed for 12 August DRDC-RDDC-2016-R112 vii

12 List of tables Table 1: Equipment locations for modems and recorder Table 2: Weather station locations viii DRDC-RDDC-2016-R112

13 1 Background The Bedford Basin, located within the Halifax Regional Municipality, Nova Scotia, is a large (approximately 8 km by 5 km) enclosed salt-water bay with a maximum depth of slightly over 70 m. The Basin has a very active acoustic environment, with significant shipping traffic and other nearby anthropogenic noise sources, including a rail yard in immediate proximity. The Basin is used regularly by the Defence Research and Development Canada (DRDC) Atlantic Research Centre for underwater acoustics experiments, particularly at the Acoustic Calibration Barge. It is also used as a training area by the Royal Canadian Navy (RCN) for diving and mine detection exercises. Because of the regular use of the Basin as an area for underwater acoustics testing, it is important to understand the nature of its underwater ambient noise. To this end, a set of ambient noise data and environmental data was acquired during the period between 24 July 2014 and 13 August Interestingly, we were unable to find any literature with previous measurements of ambient noise for the Bedford Basin, despite several decades of acoustic tests undertaken at the DRDC Acoustic Calibration Barge. The ambient noise data gathered over the course of this trial was used to test several ambient noise models for an effective model of the noise in the Bedford Basin. Along with this task, a communications experiment was conducted with underwater acoustic modems. The communications and noise trial used four modems and an acoustic recording system (the SM2M recorder, or Whalesong), which were deployed in the Bedford Basin. Signal to noise ratios and bit error rates for inter-modem communications were tested with various settings [1]. The recorder was deployed near the bottom of the Basin. It was programmed to measure the noise levels on a fixed schedule. Along with the data acquired from the modems and Whalesong recorder, water temperature and salinity data were obtained from daily CTD (conductivity-temperature-depth) casts, and weather data were obtained from surrounding weather stations. These environmental data were then used in modelling the underwater ambient noise in the Basin. 2 Experiment 2.1 Experimental setup The Bedford Basin is a silled salt-water bay within the Halifax Regional Municipality, in Nova Scotia, Canada, with a maximum depth slightly over 70 m in the centre of the basin. The floor of the basin is mostly covered in organic-rich mud in areas over 20 m deep [2], with muddy sand near the head of the basin and silty mud in the deeper parts of the basin and near the southern end [3]. Shallower, near-shore areas of the basin are primarily larger gravel, with some areas of bedrock [2]. The temperature and sound speed profiles found in the Basin are discussed in Section Figure 1 shows an aerial image (from Google Earth) of the Basin and surrounding area with overlaid bathymetry, obtained from the Ocean Mapping Group of the Geological Survey of Canada, Atlantic Geosciences Centre. DRDC-RDDC-2016-R112 1

14 Figure 1: Bathymetry of the Bedford Basin and locations of four modem nodes (3, 5, 6, 7) and Whalesong recorder. Figure 1 also shows the location of the four underwater modem nodes, each comprised of a GPS receiver, a battery and a Teledyne Benthos ATM-925 underwater acoustic modem (pictured in Figure 2), and the location of the single Wildlife Acoustics SM2M (Whalesong) recorder (also pictured in Figure 2) deployed in the Bedford Basin. The modems attached to each node operated in the 9 14 khz frequency band, and were located 5 m below the surface of the water. The Whalesong recorder was moored 2 m above the bottom of the basin in a water depth of 71 m. The recorder hydrophone has a sensitivity of ± 1 db re 1 V/μPa (inclusive of pre-amp) up to a frequency of 30 khz. The recorder sampled at 96 khz, thus making the Nyquist frequency 48 khz. The recorder has two channels recorded from the same hydrophone the left channel (Channel 1) had a 3 Hz, two-pole Butterworth high-pass filter imposed, while the right channel (Channel 2) had a 180 Hz high-pass filter and an additional 30 db of gain applied. The locations of the modems and recorder, and the water depth at those locations, are given in Table 1. 2 DRDC-RDDC-2016-R112

15 Figure 2: Teledyne Benthos modem (top) and SM2M (Whalesong) recorder (bottom). Table 1: Equipment locations for modems and recorder. Equipment Latitude ( N) Longitude ( W) Water depth (m) Node Node Node Node 7* Whalesong *Node 7 was moved to N, W on 1 August DRDC-RDDC-2016-R112 3

16 2.2 Experimental procedure Modem transmissions occupying the frequency band between 9 khz and 14 khz were scheduled through the entire period of the experiment, with approximately seven transmissions per hour. One modem would transmit a signal while the other three modems would record the signal sent and internally calculate signal-to-noise ratios (SNR). The source modem, bit rate, message length, and power level of the transmissions were varied for each signal transmission [1]. The Whalesong recorder was programmed to measure noise levels with different schedules on alternate days. Noise was recorded on a total of 20 days. On the first day (24 July 2014) and subsequent alternating (odd) days, the entire period between 21:30 and 23:30 was recorded in a 92 minute file and a subsequent 28 minute file (files were split due to restrictions on file size in the recorder computer operating system). The even day schedule, beginning on 25 July 2014, was recording two-minute periods every 30 minutes from 00:00 to 21:30 inclusive. Times are Atlantic Standard Time (offset by one hour from local due to daylight savings). The CTD data were measured using a YSI CastAway unit. The CastAway unit was deployed in different locations in the Basin twice a day for the duration of the experiment, where it was used to acquire water temperature and salinity profiles. In addition, an ADCP (Acoustic Doppler Current Profiler) was deployed during the experiment to measure currents, but the data from it has not been analyzed. 3 Data analysis 3.1 Environmental data Environmental data includes the data collected regarding the underwater environment, primarily temperature and salinity data, as well as weather data, including wind and precipitation Underwater environment The CTD data gathered by the CastAway units during the experiment were analyzed to understand the sound speed, temperature and salinity profiles in the Bedford Basin. The profiles were gathered at different spots in the Basin. A total of 99 profiles were collected during the experiment. The temperature profiles are shown in Figure 3. The top 20 m shows a pronounced thermocline, with temperatures dropping from C at the surface to 4 C at m depth. Although overall there is little change during the course of the trial, the warm layer increases in depth as the trial progresses, and later profiles (purple) show a smaller temperature gradient between about 10 m and 30 m of depth. The salinity profiles are 4 DRDC-RDDC-2016-R112

17 0 20 Depth (m) Temperature ( C) Figure 3: Temperature profiles in the Bedford Basin during the period 24 July 13 August The colour of the profile changes from red at the beginning of the trial through the rainbow to purple at the end of the trial. shown in Figure 4. The salinity in the deeper part of the basin (below 20 m) is fairly constant at a little above 31 psu, with the top few metres showing some freshening due to rainfall events and to inflow from the Sackville River. The sound speed may be calculated from the temperature, salinity, and depth using the standard Chen-Millero [4] formulation. Figure 5 shows the sound speed profiles during the course of the experiment. The sound speed profiles are uniformly downward refracting, primarily due to the warmer surface temperatures. DRDC-RDDC-2016-R112 5

18 0 20 Depth (m) Salinity (psu) Figure 4: Salinity profiles in the Bedford Basin during the period 24 July 13 August The colour of the profile changes from red at the beginning of the trial through the rainbow to purple at the end of the trial. 6 DRDC-RDDC-2016-R112

19 0 20 Depth (m) Sound speed (m/s) Figure 5: Sound speed profiles in the Bedford Basin during the period 24 July 13 August The colour of the profile changes from red at the beginning of the trial through the rainbow to purple at the end of the trial. DRDC-RDDC-2016-R112 7

20 3.1.2 Weather conditions To aid in modelling the ambient noise data, estimates of wind speed were required. Six weather stations were located within 25 km of the recorder, from which data were taken. The station locations are given in Table 2. Table 2: Weather station locations. Station Latitude ( N) Longitude ( W) Bedford Basin Bedford Range Shearwater A Shearwater Jetty Shearwater RCS Windsor Temperature, wind speed and relative humidity data were acquired from each of the weather stations, and the data were interpolated to the location of the recorder. Several interpolation methods included with the IDL (Interactive Data Language) programming package were tested, with no significant differences noted between them. Given that, the default method of using inverse distance weighting was chosen for interpolation. Only the wind speed was used for modelling. Reported wind speeds are the average wind speed over 2 minutes, recorded hourly, at 10 m above the surface. This averaged wind speed is shown in Figure 6. Precipitation data were acquired from the Halifax Airport weather station (approximately 20 km away no closer station had good precipitation data). Figure 7 plots precipitation levels over the experimental period, with 1 corresponding to drizzle, 2 to rain showers, and 3 to rain. 8 DRDC-RDDC-2016-R112

21 40 30 Wind speed (m/s) /21 07/26 07/31 08/05 08/10 08/15 Date Figure 6: Wind speeds during the period 24 July 13 August Figure 7: Relative precipitation levels during the period 24 July 13 August DRDC-RDDC-2016-R112 9

22 3.2 Ambient noise data Power spectral levels The power spectral density (PSD) of the recorded noise was determined over the course of the experiment using Welch s method [5], where periodograms were generated for each one-minute segment of data, using a point FFT (1 Hz resolution) with 50% overlap and a Hanning window. This gave PSDs for each of 2092 one-minute periods. The mean PSD and the spectra of percentiles 5, 25, 50 (median), 75, and 95 were also computed (i.e., the spectral levels that were exceeded 5%, 25%, 50%, 75% and 95% of the time). These spectra are shown in Figure 8. The mean sound pressure level (SPL) in each third-octave band is also shown, along with the minimum, maximum, median, and 25 and 75 percentiles. Figure 8: Top: Mean power spectrum and percentiles for Channel 1 over the entire experiment. Bottom: Mean and percentiles of third-octave sound pressure level (black line). 10 DRDC-RDDC-2016-R112

23 It is evident both from the high mean levels relative to the median and from the onesided nature of the third-octave minima and maxima that the power spectra are strongly influenced by occasional very high noise levels, in all bands. These levels are less frequent than the 5% occurrence values, which at low frequencies (below 300 Hz) and high frequencies (above 20 khz) are less than the mean. It is possible that these infrequent very loud events are nearby shipping. There is also a band of significant noise that appears between 9 khz and 14 khz. This noise is generated by the modem in this experiment and it occupies the 5th percentile because when there is a (relatively infrequent) transmission it is significantly louder than the ambient noise levels. There also appear to be strong harmonics of 12 Hz and of 750 Hz, strong enough to be seen in the median spectrum (and even occasionally in the 25% spectrum). In Figure 9, the spectra for the second channel, with the additional gain and the 180 Hz high-pass filter, are plotted. In this case, the harmonics are reduced with respect to the ambient noise, and are therefore likely to be electronic noise that occurs after the pre-amp stage. A comparison of Figure 8 and Figure 9 shows differences between the channels beyond the initial slope caused by the high-pass filters. Figure 10 plots the mean difference between the spectra measured using the two channels. There is a significant difference caused by the filters as would be expected, although it is apparent that the filters both have a less steep roll-off than desired (closer to 15 db/decade rather than the 40 db/decade expected), and that there are substantial ripples at low frequencies. An adjustment of 1.1 db was made to the difference, as the difference between channels was 28.9 db rather than the 30 db expected from the nominal gain difference. There are a few large (3 db) differences between channels at higher frequencies, potentially due to high transients causing saturation in Channel Ambient noise time dependence The ambient noise power spectra shown in Section are based on the entire period of the experiment. Other information can be obtained by looking at the spectrograms of the noise, i.e., the noise power as a function of both frequency and time. Figure 11 shows the spectrogram for the noise recorded on Channel 2 of the Whalesong recorder on 25 July The power spectral densities for each of the two-minute recordings are plotted as a function of time of day and frequency (on a logarithmic scale). The wind speed at the recorder is also plotted. Annex A has spectrograms for the remainder of the experiment. DRDC-RDDC-2016-R112 11

24 Figure 9: Top: Mean power spectrum and percentiles for Channel 2 over the entire experiment. Bottom: Mean and percentiles of third-octave sound pressure level (black line). 12 DRDC-RDDC-2016-R112

25 Figure 10: Channel noise level difference. The black line shows the theoretical effect of the two-pole Butterworth filters at 3 Hz and 180 Hz, and the red line shows the average spectrum of Channel 2 minus Channel 1. An adjustment of 1.1 db has been added to the difference. DRDC-RDDC-2016-R112 13

26 Figure 11: Ambient noise spectrogram (Channel 2) and wind speed (black line) on 25 July The logarithmic frequency is given on the left y-axis, and the wind speed in m/s on the right y-axis. The x-axis shows the time of day in hours. 14 DRDC-RDDC-2016-R112

27 In Figure 11, we can see that the power spectral density varies substantially as a function of time, particularly in the frequency range from about 100 Hz to about 1 khz, with peaks near 100 db re 1 µpa 2 /Hz in that frequency range at times from AST and AST (early morning and evening). The dependence of ambient noise on wind speed will be investigated via noise models in Section 4.1, but it is not obvious from Figure 11, that there is a correlation between noise and wind speed in any given frequency band. The two models that will be discussed have a linear dependence between the logarithm of the wind speed and the noise level. To investigate possible correlations, the third-octave SPL values for time periods corresponding to the wind measurement time periods (i.e., the first two minutes of each one-hour interval) were plotted against the logarithm of the wind speed and the Pearson correlation coefficient was calculated for each third-octave band. An example of this is shown in Figure 12, for the band Hz, which has the strongest correlation between noise and wind speed. The wind speeds cover the range of nearly calm to almost 40 m/s. The minimum measurable noise level, corresponding to the flat part of the L 95 spectrum in Figure 8 (top), is about 73 db. This gives the floor of the SPL in Figure 12. The plot shows a very weak degree of correlation, with an R 2 value of Figure 12: Third-octave SPL for the band 4064 Hz to 5120 Hz vs. measured wind speed. Green line shows the least-squares fit between the logarithm of the wind speed and the SPL. Figure 13 shows the correlation coefficient for all third-octave bands (SPL vs. log base 10 of wind speed). Correlations are overall weak nearly zero outside the 1 10 khz frequency band. DRDC-RDDC-2016-R112 15

28 Figure 13: Pearson correlation coefficient R 2 vs. frequency for third-octave SPL correlation with the logarithm of wind speed Noise statistics It is clear from the results shown in Section that the ambient noise sound pressure levels are not Gaussian-distributed. To investigate the statistics of the noise, we computed the sound pressure levels (SPL) in each third-octave bin by integrating the power spectral density over the frequencies in that bin, and then calculated the histograms for the SPL over the course of the experiment, using these to estimate the probability density function. An example for the third-octave bin 806 Hz to 1016 Hz is shown in Figure 14. In Figure 14, it is apparent that for this frequency band, the probability density is skewed to the right (higher SPL levels). The skewness and excess kurtosis (i.e., kurtosis minus three) for each third-octave band were computed and are plotted in Figure 15. From Figure 15, we can see that the skewness of the distributions for all of the thirdoctave bands is positive, indicating that there is a higher incidence of higher noise power events compared to lower noise power events. This is unsurprising, as there is a minimum measurable ambient noise threshold. Below 100 Hz and above 10 khz, the skewness is quite a bit higher than through the mid-frequency bands. The distribution is also platykurtic, or flattened, at those frequencies. This tends to imply that the distribution is somewhat more uniform in those frequency bands. Above 10 khz, the kurtosis is large and positive. This, and the strongly positive skewness, is probably due to the effects of the modem transmissions, which introduce an abnormally large number of high noise level events to the probability distribution. 16 DRDC-RDDC-2016-R112

29 Figure 14: Histogram of SPL for third-octave band 806 Hz to 1016 Hz Skewness Kurtosis Skewness Kurtosis Frequency (Hz) 0.0 Figure 15: Skewness and excess kurtosis vs. frequency for third-octave bands. DRDC-RDDC-2016-R112 17

30 4 Noise modelling 4.1 Ambient noise models Underwater ambient noise is the result of a combination of shipping, other commercial activities, biologics, and wind and precipitation. In the absence of shipping and biological activity, the ambient noise typically depends on wind speed from frequencies of a few tens of Hz to a few khz. Heavy ship traffic can not only increase ambient noise (typically in the frequency range of hundreds of Hz) but the resulting noise can be highly directional. There are a number of historical reviews of underwater ambient noise in the literature (e.g., [6, 7, 8]). Wenz [7] developed a set of ambient noise spectral curves for deep water and shallow water based on observations, with shallow water levels that are in general several decibels higher than the corresponding deep-water levels for the same frequencies and wind speeds. At the very lowest frequencies, oceanic turbulence and seismic background noise are the dominant noise sources, while higher than 100 khz, thermal noise increases rapidly with frequency [6]. Finally, shipping noise tends to occupy the frequency band near 100 Hz curves based on observed levels have been created [9, 7]. Figure 16 shows example power spectra using noise curves for light and heavy shipping in deep and shallow water, turbulence, and thermal noise. The power spectra for wind noise are computed using the Merklinger-Stockhausen model [10] for 5 knot and 40 knot winds. The shallow water spectra are higher than the deep water by 3 db [11, 12]. Intermittent sources of noise, such as precipitation, ice-cracking, and biologic, are not shown in this figure. For the purposes of this report, two different ambient noise models were used for comparison to the data. The models tested were the Merklinger-Stockhausen model [10], with the addition of a turbulence term and a shipping term, and an ambient noise model taken from the APL-UW (Applied Physics Laboratory, University of Washington) Handbook, based on [13]. The APL-UW model is solely wind-speed based, and is in fact a high-frequency noise model for noise at the air/sea interface caused by wind at normal incidence. The APL-UW Handbook suggests that the frequency range of applicability of this model is between 1 khz and 50 khz in the absence of rain or biological noise. The inclusion of the turbulence and shipping noise into the Merklinger-Stockhausen model extends the range of applicability to include frequencies down to a few Hertz. Figure 17 shows an example comparing the two models. The Merklinger-Stockhausen model, as used, consists of a shipping curve based on the NUSC (Naval Underwater Systems Center) shipping level model, a turbulence curve, and a wind speed curve (labelled Merk-Stock), which includes the 3 db of extra noise experimentally found in shallow water [11, 12]. The ambient noise level L (in decibels) for the wind speed curve 18 DRDC-RDDC-2016-R112

31 µ Figure 16: Typical ambient noise spectra from turbulence, molecular agitation (thermal), shipping, and wind. DRDC-RDDC-2016-R112 19

32 µ Figure 17: Example comparison of both models using a NUSC shipping level of 4 and a wind speed of 10 m/s. For the APL-UW model, no temperature difference between sea surface and air is used. 20 DRDC-RDDC-2016-R112

33 is given by: L = L 0 = 25 f L log log 10 U 25 1/25 f L 0 17 log 10, log 10 U log10 U log 10 U 17 log 10, 770 (1) where U is the wind speed in knots and f is the frequency in Hz. To this curve was added the noise based on shipping level, which was computed by matching the NUSC model noise level with the measured median noise level between 50 Hz and 60 Hz during the two-minute time period that wind measurements were taken. Figure 18 shows the resulting plot of median noise level and corresponding NUSC shipping level parameter for those periods. Most of the first, third, etc. days are not shown as noise was only recorded between 21:30 and 23:30 on those days. Figure 18: Median ambient noise levels in Hz band and corresponding NUSC shipping levels. The version of the APL-UW model used is the surface noise power spectral density, given an omnidirectional receiver and no contribution from multi-paths (i.e., the assumption is that of a soft bottom with high reflection losses). Given these assumptions, the noise PSD level L in decibels is dependent solely on wind speed and frequency, and is given by: L = 10log 10 π + A, (2) DRDC-RDDC-2016-R112 21

34 where A is the noise source level (in decibels) at the air/sea interface at vertical incidence, with respect to a reference value at a frequency of 20 khz dependent only upon wind speed: A = A log 10 f khz, A 20 = log 10 U, T < 1 C = log 10 U 0.26( T 1) 2, T 1 C. (3) Here f khz is the frequency in khz, U is the wind speed in m/s at a height of 10 m, and T is the air temperature minus the sea surface temperature. 4.2 Model-data comparisons Figure 19 shows the difference between the ambient noise data at each frequency and the Merklinger-Stockhausen model results calculated using the methodology given in Section 4.1, while Figure 20 shows the data minus the model results for the APL-UW model. In both cases, ambient noise data from the two-minute window where the wind speed was measured were used for the comparison. Therefore, 24 July and subsequent alternate days only have data for the last two hours of each day shown. 22 DRDC-RDDC-2016-R112

35 Figure 19: Ambient noise data minus Merklinger-Stockhausen model results with respect to frequency for period of the experiment. DRDC-RDDC-2016-R112 23

36 Figure 20: Ambient noise data minus APL-UW model results with respect to frequency for period of the experiment. 24 DRDC-RDDC-2016-R112

37 Both sets of model-data difference show considerably higher measured ambient noise across most of the frequency band for the period of 1 August to 8 August (particularly for 6 August), while the Merklinger-Stockhausen model overestimates noise in the band from 100 Hz to about 3 khz particularly on 31 July and 10 August. Modem transmissions show up clearly in the frequency region near 10 khz. The model-data differences below about 50 Hz for the Merklinger-Stockhausen model show more low frequency noise than modelled, possibly due to inadequate modelling of the effects of the low-frequency filter. For the APL-UW model, the model tends to predict a higher ambient noise level than that shown in the data outside the period 1 August to 8 August, across all frequencies. That is unsurprising in the lower frequencies, as there is no real attempt to model the low-frequency noise beyond a continuation of the wind noise curve. However, the model also overestimates noise at frequencies above 1 khz, where the wind noise curve would be expected to work. This may be due to the fact that the APL-UW model is premised on fully developed surface waves, as opposed to enclosed water such as the Basin. The noise data were also compared to the precipitation data shown in Figure 7. There was no obvious correlation between precipitation levels and the higher ambient noise data compared to model results. Another way of looking at the model-data difference is to see if there is a noticeable trend by time of day. Figure 21 shows the median results of the Merklinger-Stockhausen model predictions subtracted from the ambient noise data for frequency and time of day. Again, at low frequencies (less than about 70 Hz) the model under-predicts the data, possibly due to the filter correction. Overall, the less than 1 khz frequency band is under-predicted by the model at night, while the data is over-predicted by the model from about 1 3 khz, particularly in the early evening. The higher noise levels above 10 khz in the afternoon are due to the modem transmissions. Figure 22 shows the equivalent model-data difference, using the APL-UW model. It is clear that in general the model over-predicts the ambient noise for most frequencies, except for the night-time and for higher frequencies (above 10 khz) during the afternoons, which includes the modem transmissions. A direct comparison of the model-data mismatches is shown in Figure 23. Except at frequencies below 30 Hz, the Merklinger-Stockhausen model is a better fit to the data, with the mean absolute mismatch over third-octave bands through the experiment typically less than 5 db. The standard deviation is quite high, between 8 db and 12 db from 100 Hz to 10 khz. The APL-UW mismatch is however higher for all frequency bands. DRDC-RDDC-2016-R112 25

38 Figure 21: Difference spectrogram between the data and Merklinger-Stockhausen model vs. time of day. Figure 22: Difference spectrogram between the data and APL-UW model vs. time of day. 26 DRDC-RDDC-2016-R112

39 Figure 23: Third-octave mean of data minus model for Merklinger-Stockhausen model and APL-UW model. Error bars show one standard deviation. DRDC-RDDC-2016-R112 27

40 5 Conclusions Underwater ambient noise was measured during a sea trial that was carried out in the Bedford Basin (NS, Canada) in the summer of 2014, while a set of modems were used to send messages for a communications intelligibility experiment. The power spectral density and statistics of the recorded noise were determined over the period of the experiment, and the noise was shown to be non-gaussian. The ambient noise power spectra were then compared to existing ambient noise models, using weather data acquired during the experiment. The model-data agreement for each comparison varies significantly by time of day and frequency band. In general, the Merklinger-Stockhausen model was more accurate than an extension of the wind-noise part of the APL-UW model. However, there were fairly large discrepancies between the model and data even on an averaged basis. In the low frequency regime, where wind noise does not dominate, the APL-UW model is not expected to be representative, and in fact the model over-predicts ambient noise levels for most frequency bands over much of the experiment, although with reasonable agreement in the later part of the day and for certain dates in the band from 5 to 20 khz. The model in general over-predicts noise at frequencies above 1 khz, where the wind noise curve would be expected to work. This may be due to the fact that the APL-UW model is premised on fully developed surface waves, as opposed to enclosed water such as the Basin. The Merklinger-Stockhausen model under-predicts low frequency (under 60 Hz) ambient noise, possibly due to inadequate modelling of the effects of the low-frequency filter, or possibly due to high levels of mechanical noise (including the nearby port and railways). It also under-predicts the measured ambient noise across most of the frequency band for the period of 1 August to 8 August (particularly for 6 August), and over-predicts noise in the band from 100 Hz to about 3 khz particularly on 31 July and 10 August. However, the median differences through the course of the experiment are fairly accurate. It is likely that nearby shipping, mechanical, and other anthropogenically generated noise sources all contribute to the model-data differences, for example, the higher noise levels above 10 khz in the afternoon are due to modem transmissions. Overall, neither of the noise models tested provides a complete picture of the ambient underwater noise in the Bedford Basin. However, the modified Merklinger-Stockhausen model can set some useful bounds on expected noise levels, and the measured noise data set will also help inform the requirements for equipment testing in the Basin. 28 DRDC-RDDC-2016-R112

41 References [1] Vandenberg, D. and Blouin, S. ( ), Acoustic Modem Parameters Data Analysis. Private Communication. [2] Fader, G. B. and Buckley, D. E. (1995), Environmental geology of Halifax Harbour, Nova Scotia, Geoscience Canada, 22(4), [3] Miller, A., Mudie, P., and Scott, D. B. (1982), Holocene history of Bedford Basin, Nova Scotia: foraminifera, dinoflagellate, and pollen records, Canadian Journal of Earth Sciences, 19(12), [4] Chen, C. and Millero, F. (1977), Speed of sound in seawater at high pressures, The Journal of the Acoustical Society of America, 62(5), [5] Welch, P. D. (1967), The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms, IEEE Transactions on Audio and Electroacoustics, AU-15(2), [6] Urick, R. (1984), Ambient noise in the sea, Undersea Warfare Technology Office, Naval Sea Systems Command, Dept. of the Navy. [7] Wenz, G. (1962), Acoustic ambient noise in the ocean: spectra and sources, The Journal of the Acoustical Society of America, 34(12), [8] Zakarauskas, P. (1986), Ambient noise in shallow water: a literature review, Canadian Acoustics, 14(3), [9] Sadowski, W., Katz, R., and McFadden, K. (1984), Ambient noise standards for acoustic modeling and analysis, (NUSC Technical Document 7265 (3)) Naval Underwater Systems Center. [10] Mercklinger, H. M. and Stockhausen, J. H. (1979), Formulae for estimation of undersea noise spectra, Journal of the Acoustical Society of America, 65, S88. [11] Hazen, M. and Desharnais, F. (1997), The Eastern Canada Shallow Water Ambient Noise experiment, In OCEANS 97 MTS/IEEE Conference Proceedings, Vol. 1, pp [12] Piggott, C. (1964), Ambient sea noise at low frequencies in shallow water of the Scotian Shelf, The Journal of the Acoustical Society of America, 36, [13] Jackson, D. (1994), APL-UW high-frequency ocean environmental acoustic models handbook, (Technical Report 9407) Applied Physics Laboratory, University of Washington. DRDC-RDDC-2016-R112 29

42 This page intentionally left blank. 30 DRDC-RDDC-2016-R112

43 Annex A: Ambient noise spectrograms In this annex the spectrograms for the ambient noise recorded on Channel 1 of the Whalesong recorder are plotted along with the measured wind speed. The spectrograms show the power spectral densities (PSDs) for each of the two-minute recordings as a function of time of day and frequency (on a logarithmic scale). Note that the scales of both the PSD levels and the wind speed vary between plots. DRDC-RDDC-2016-R112 31

44 Figure A.1: Ambient noise spectrogram and wind speed for 24 July DRDC-RDDC-2016-R112

45 Figure A.2: Ambient noise spectrogram and wind speed for 25 July DRDC-RDDC-2016-R112 33

46 Figure A.3: Ambient noise spectrogram and wind speed for 26 July DRDC-RDDC-2016-R112

47 Figure A.4: Ambient noise spectrogram and wind speed for 27 July DRDC-RDDC-2016-R112 35

48 Figure A.5: Ambient noise spectrogram and wind speed for 28 July DRDC-RDDC-2016-R112

49 Figure A.6: Ambient noise spectrogram and wind speed for 29 July DRDC-RDDC-2016-R112 37

50 Figure A.7: Ambient noise spectrogram and wind speed for 30 July DRDC-RDDC-2016-R112

51 Figure A.8: Ambient noise spectrogram and wind speed for 31 July DRDC-RDDC-2016-R112 39

52 Figure A.9: Ambient noise spectrogram and wind speed for 1 August DRDC-RDDC-2016-R112

53 Figure A.10: Ambient noise spectrogram and wind speed for 2 August DRDC-RDDC-2016-R112 41

54 Figure A.11: Ambient noise spectrogram and wind speed for 3 August DRDC-RDDC-2016-R112

55 Figure A.12: Ambient noise spectrogram and wind speed for 4 August DRDC-RDDC-2016-R112 43

56 Figure A.13: Ambient noise spectrogram and wind speed for 5 August DRDC-RDDC-2016-R112

57 Figure A.14: Ambient noise spectrogram and wind speed for 6 August DRDC-RDDC-2016-R112 45

58 Figure A.15: Ambient noise spectrogram and wind speed for 7 August DRDC-RDDC-2016-R112

59 Figure A.16: Ambient noise spectrogram and wind speed for 8 August DRDC-RDDC-2016-R112 47

60 Figure A.17: Ambient noise spectrogram and wind speed for 9 August DRDC-RDDC-2016-R112

61 Figure A.18: Ambient noise spectrogram and wind speed for 10 August DRDC-RDDC-2016-R112 49

62 Figure A.19: Ambient noise spectrogram and wind speed for 11 August DRDC-RDDC-2016-R112

63 Figure A.20: Ambient noise spectrogram and wind speed for 12 August DRDC-RDDC-2016-R112 51

64 This page intentionally left blank. 52 DRDC-RDDC-2016-R112

65 DOCUMENT CONTROL DATA (Security markings for the title, abstract and indexing annotation must be entered when the document is Classified or Designated.) 1. ORIGINATOR (The name and address of the organization preparing the document. Organizations for whom the document was prepared, e.g. Centre sponsoring a contractor s report, or tasking agency, are entered in section 8.) DRDC Atlantic Research Centre PO Box 1012, Dartmouth NS B2Y 3Z7, Canada 2a. SECURITY MARKING (Overall security marking of the document, including supplemental markings if applicable.) UNCLASSIFIED 2b. CONTROLLED GOODS (NON-CONTROLLED GOODS) DMC A REVIEW: GCEC DECEMBER TITLE (The complete document title as indicated on the title page. Its classification should be indicated by the appropriate abbreviation (S, C or U) in parentheses after the title.) Ambient noise in the Bedford Basin 4. AUTHORS (Last name, followed by initials ranks, titles, etc. not to be used.) Chattopadhyay, K.; Pecknold, S. P. 5. DATE OF PUBLICATION (Month and year of publication of document.) August a. NO. OF PAGES (Total containing information. Include Annexes, Appendices, etc.) 60 6b. NO. OF REFS (Total cited in document.) DESCRIPTIVE NOTES (The category of the document, e.g. technical report, technical note or memorandum. If appropriate, enter the type of report, e.g. interim, progress, summary, annual or final. Give the inclusive dates when a specific reporting period is covered.) Scientific Report 8. SPONSORING ACTIVITY (The name of the department project office or laboratory sponsoring the research and development include address.) DRDC Atlantic Research Centre PO Box 1012, Dartmouth NS B2Y 3Z7, Canada 9a. PROJECT OR GRANT NO. (If appropriate, the applicable research and development project or grant number under which the document was written. Please specify whether project or grant.) 01CA 10a. ORIGINATOR S DOCUMENT NUMBER (The official document number by which the document is identified by the originating activity. This number must be unique to this document.) DRDC-RDDC-2016-R112 9b. CONTRACT NO. (If appropriate, the applicable number under which the document was written.) 10b. OTHER DOCUMENT NO(s). (Any other numbers which may be assigned this document either by the originator or by the sponsor.) 11. DOCUMENT AVAILABILITY (Any limitations on further dissemination of the document, other than those imposed by security classification.) ( X ) Unlimited distribution ( ) Defence departments and defence contractors; further distribution only as approved ( ) Defence departments and Canadian defence contractors; further distribution only as approved ( ) Government departments and agencies; further distribution only as approved ( ) Defence departments; further distribution only as approved ( ) Other (please specify): 12. DOCUMENT ANNOUNCEMENT (Any limitation to the bibliographic announcement of this document. This will normally correspond to the Document Availability (11). However, where further distribution (beyond the audience specified in (11)) is possible, a wider announcement audience may be selected.) Unlimited

66 13. ABSTRACT (A brief and factual summary of the document. It may also appear elsewhere in the body of the document itself. It is highly desirable that the abstract of classified documents be unclassified. Each paragraph of the abstract shall begin with an indication of the security classification of the information in the paragraph (unless the document itself is unclassified) represented as (S), (C), or (U). It is not necessary to include here abstracts in both official languages unless the text is bilingual.) To facilitate the development of better underwater communications technology, a sea trial was carried out in the Bedford Basin (NS, Canada) in the summer of Modems were used to send messages and coincident ambient noise measurements were recorded. The ambient noise data were analyzed in order to verify the signal to noise ratios of the equipment and to investigate the characteristics of the Bedford Basin noise during this time. Specifically, the ambient noise power spectra were compared to two existing ambient noise models. The model-data agreement for each comparison varies significantly by time of day and frequency band. In general, the Merklinger-Stockhausen model was more accurate than an extension of the wind-noise part of the APL-UW model. However, there were fairly large discrepancies between the model and data even on an averaged basis. 14. KEYWORDS, DESCRIPTORS or IDENTIFIERS (Technically meaningful terms or short phrases that characterize a document and could be helpful in cataloguing the document. They should be selected so that no security classification is required. Identifiers, such as equipment model designation, trade name, military project code name, geographic location may also be included. If possible keywords should be selected from a published thesaurus. e.g. Thesaurus of Engineering and Scientific Terms (TEST) and that thesaurus identified. If it is not possible to select indexing terms which are Unclassified, the classification of each should be indicated as with the title.) DRDC Scientific Report; ambient noise