OS 72B-0355 Analysis of Acoustic Signals from Ship Traffic at Pioneer Seamount
Carl O. Vuosalo,1 Craig Huber,1 Michael D. Hoffman,1 Newell Garfield,2 and Roger Bland1
1. Physics and Astronomy Department and Romberg Tiburon Center for Environmental Studies, San Francisco State University
2. Department of Geosciences and Romberg Tiburon Center for Environmental Studies, San Francisco State University
In August 2001 a vertical linear array (VLA) of four hydrophones was installed by NOAA-PMEL at Pioneer Seamount, 95 km off cable
the California coast and 930 m below the surface. The four channels, digitized at 1000 Hz with 16-bit precision, are available in cable BREAK
near real time. In this data set the loudest and most obvious signals are those created by passing ships. We present a method for
analyzing the ship signals to determine the speed of each ship and its distance of closest approach to the array. The observatory is
now offline due to a broken cable, and due to uncertain funding, no repairs have yet been planned.
Ship Detection and Speed Calculation
Software was developed in the IDL programming language to automatically
ROMBERG TIBURON CENTER
find ship signals in the data set and then calculate the speed and distance of closest
approach of each ship. The first step in this process is to make a Fourier transform
spectrogram of the raw acoustic data. This transformed data is scanned for the
Topography in the vicinity of the array, which is shown distinctive interference pattern created by ships. These signatures are considered
as three vertical dots. to be two sharp peaks that are integer multiples of each other and that occur below
21 Hz. The second step is to perform a Fourier transform on the Fourier transform
spectrograms. At those locations previously found to contain ship signatures, the
double Fourier transforms are scanned to find a sequence of sharp peaks. This set
of peaks is then fitted with a calculated distance and time curve, and if a
reasonable fit is achieved, the presence of a ship is confirmed, and its speed and
distance can be obtained from the parameters of the curve. Fourier transforms of two ship signals to give frequency vs. time spectrograms. The top panels
show data summed from all four hydrophones. The middle panels show data from only the top
hydrophone of the array. The bottom panels show Fourier transforms of the middle panels.
Hz FOUR HYDROPHONES
FOUR HYDROPHONES COMPUTER SIMULATION
• dmon = 3600 m
R • speed 15.8 m/s (30 knots)
• flat sea bottom
q • white noise
R d / tan q
a sin q n
Underwater topography near Pioneer Seamount. (Graphics courtesy of PMEL/NOAA)
Fourier f 500
Hz ONE HYDROPHONE
spectrograms Rd af / c 2
Scan result for Faculty Talks, September 9, 2002
One method of calculating ship range R, based upon separation
Fit set of peaks to distance vs. between maxima in the Fourier transform f, separation between
time curve to confirm ship and hydrophones a, and the speed of sound in water c.
determine speed and distance. 0
0 The computer simulation shown to the right, for the faster ship, “FFT OF FFT”
“FFT OF FFT”
9 m/s 14 m/s (28 knots) 7 m/s 9 m/s 6 m/s is based on this geometry, using an expression for the intensity,
THE LOUDEST SHIP SIGNALS. Signals like this arrive about once every three days. 6:00-
I 2 cos 2
The are the loudest sounds observed at Pioneer Seamount. UT r
where is the difference in phase for paths from the ship to a
10:00- single hydrophone and to its image beneath the sea floor. For Fitted speed curves for two ships. Plot of surface range in meters from the array vs. time in seconds.
UT the simulation shown and for the spectrogram to which it is Green curves based upon blue points. Red points were rejected.
14:00 compared, only the signal from the upper of the four
hydrophones was used.
13 m/s 9 m/s 20:00-
16 m/s (32 knots) 22:00
SOME FASTER SHIPS. The speed of these ships has been estimated from the slope of the UT
interference maxima, since the fitting procedure described to the right is disrupted by the presence of
the constant-frequency lines on these spectrograms (vertical lines). These lines are probably due to Spectrograms for an 18-hour period (23:00 July 20 to 17:00 July 21, PDT)
engine noise, while the smoother part of the pattern is due to prop noise. showing numerous ship signals. The patterns gating on and off probably
represent boats stopping or idling while fishing. (All spectrograms 0-500
Hz, frequency increasing upwards.)
Ship at 08:15 on Sept. 8, 2001. Ship at 21:15 on Sept. 10, 2001.
Help and advice is gratefully acknowledged from:Chris Fox, Jonathan Klay, Andy Lau, and Haru Matsumoto, NOAA-PMEL; Jim Mercer Speed: 6.1 m/s. Closest approach: 645 m. Speed: 15.8 m/s. Closest approach: 3656 m.
and Lyle Gullings, APL, University of Washington; and John Bourg and Jim Lockhart, San Francisco State University This calculated closest approach may be