Long-Term Ambient Noise Statistics in the Gulf of Mexico Mark A. Snyder & Peter A. Orlin Naval Oceanographic Office Stennis Space Center, MS Anthony I. Eller Science Applications International Corporation
Jan 14, 2016
Long-Term Ambient NoiseStatistics
in theGulf of Mexico
Mark A. Snyder & Peter A. Orlin
Naval Oceanographic Office
Stennis Space Center, MS
Anthony I. Eller
Science Applications International Corporation
EARS* Data
AcousticRelease
Floats
EARSData
Logger
• Bottom-moored omni-directional hydrophone• Bandwidth of 10 Hz - 1000 Hz• 14 months of data (Apr 2004 - May 2005)• Water depth ~ 3200 meters• Hydrophone depth ~ 2935 meters• Vicinity 27.5 N, 86.1 W (about 159 nm south of Panama City, FL and 196 nm west of Tampa, FL)
* Environmental Acoustic Recording System
Location of EARS and NDBC* Weather Buoys
NDBC 42003
EARS
103 nm
89 nm
3200 m
3200 m
55 m
NDBC 42036
* National Data Buoy Center
Monthly
Trends
Monthly Statistics*
• Mean
• Median
• Standard deviation
• Skewness
• Kurtosis
• Coherence time**
* For 8 third-octave bands.
** Time for autocorrelation to decay to e-1 of its zero-lag value.
1 year cycle
Hurricanes
Hurricanes
14-Month
Statistics
Low frequency band
Positive skewness
Chi – Square PDF
Apr04 May05
Apr04 May05
High frequency bandNegative skewness
Hurricanes Winter Storms
1st order Gauss-Markov process is characterized by an exponentially-decaying autocorrelation.
Coherence time = 2.97 hours
400 Hz
Variability
Time
Scales
• Power spectrum of 14-month time series shows how the energy associated with variability is spread over long and short time scales.
• Each vertical bar = variance in each 1/10-decade* freq band.
• Sum of all vertical bars = total variance.
• Low frequency band
• Most of the variability is in time scales near 10 hours
• Red curve is plot of 1st order Gauss-Markov process
4 days 6 weeks 1 year* 1/10-decade ≈ 1/3-octave
4 days 6 weeks 1 year
• High frequency band• Most of the variability is in time scales near 100 hours
Frequency Coherence
2 octaves to left and right of center frequency have correlation coefficient ≥ 0.5
Spatial Coherence
A1 A3A6
2.29 km2.56 km
Water depth = 3200 m at all 3 sites
Hydrophone depth = 2935 m at all 3 sites
10 month comparison
• 100 Hz - more affected by local noise sources
• 1000 Hz – wind is correlated over large distances
• 100 Hz - more affected by local noise sources
• 1000 Hz – wind is correlated over large distances
Comparison to NDBC
Weather Data
• 14-month avg wind speed = 11.3 knots
• Avg significant wave height = 1.06 m
• Avg Beaufort Wind Force = 3.5
• Moderate to heavy shipping
• Shipping level 6-7 on scale of 1-9
• 14-month avg wind speed = 11.3 knots• Avg significant wave height = 1.06 m• Avg Beaufort Wind Force = 3.5• Moderate to heavy shipping• (Shipping level = 6-7 on scale of 1-9, with 1 = light, 9 = very heavy)
Best-Fit Density Functions (14 Months)
Fc (Hz)
Best Fit PDF (3 Moments)
Comments
25 Rayleigh σR = 6.06 50 Chi-Square n = 7 100 Chi-Square n = 6 200 Rayleigh σR = 5.60
σR = Rayleigh parameter.
n = degrees of freedom.
14-Month Summary• Ambient noise at low frequencies (25 – 400 Hz)
Mean > median > mode (2 – 3 dB spread)
All 3 values close and predicted by moderate to heavy shipping. Location of all 3 caused positive skewness (skewed towards peaks).
• Ambient noise at high frequencies (630 – 950 Hz)
Mode > median > mean (2 – 3 dB spread)
All 3 values close and predicted by avg BWF = 3.5 (11.3 knots avg wind). Location of all 3 caused negative skewness (skewed towards troughs).
14-Month Summary
• Coherence time was low (2 – 4 hours) in shipping bands (25 – 400 Hz)
• Coherence time was high (14 – 21 hours) in weather bands (630 – 950 Hz)
• Monthly coherence time was highest during extreme wind conditions
14-Month Summary• Temporal variability occurred over 3 time scales:
7 - 22 hours (shipping-related) 56 - 282 hours (2 - 12 days, weather-related) 8 - 12 months (1 year cycle)
• The 25 Hz time series had a strong 8-hour component (sinusoidal autocorrelation; not shipping or weather)
• The 50, 100 and 200 Hz frequency bands were fit by a 1st order Gauss-Markov process (well characterized by 3 parameters: mean, variance and coherence time)
• More complicated structure in other bands
Avg BWF = 2.5
Mean = 56.26 dB
σ = 5.78 dBRange = 38.27 dB
Skewness = 0.45
C.T. = 1.74 hours
Avg BWF = 4
Mean = 62.73 dB
σ = 4.67 dBRange = 31.81 dB
Skewness = -0.51
C.T. = 10.31 hours
Data Processing
2048 Point FFT
10 Minute Avg
Power Spectra
Separate Data
Into
14 Months
Bandpass Each Month’s
Data Over
Eight 1/3-Octave Bands
Compute Monthly
Statistics Over Each
Frequency Band
Raw Acoustic
Time Series Data
Average 732 (0.82 seconds each) periodograms.
Sampled at 2.5 kHz.
Remove disk spin and clips.
Compute the average power in each band every 10 minutes.
Δf = 1.22 Hz.
FC = 25, 50, 100, 200,
400, 630, 800, 950 Hz
Frances Ivan I JeanneIvan II