CEAREX Drift Experiment Ambient Noise Observations

                         Robert S. Pritchard
                           IceCasting, Inc.
                        11042 Sand Point Way NE
                       Seattle, WA 98125-5846 USA


1. Introduction

     CEAREX Drift Operations used the ship POLARBJORN as a scientific base. 
The ship steamed into the polar pack ice during early September 1988 with 
icebreaker support from the USCG NORTHWIND.  POLARBJORN was allowed to 
freeze into the ice on 16 September at 82 degrees 41 minutes N, 32 degrees
26 minutes E.  The ship drifted southeastward toward Viktoriya Island, then
southwestward past Kvitoya Island, and finally into the Barents Sea.  All 
drift operations were completed by mid-January 1989.  The drift experiment 
was designed to observe simultaneous atmosphere, ice, and ocean behavior,
including wind, air temperature, current, large scale ice motion, local 
ice motion, ice stress, and under-ice noise ["CEAREX drift experiment", 
Pritchard, R.S. and twenty-eight others, EOS, 71(40), 1990, p. 1115-1118]. 

     Ambient noise was measured using omni-directional hydrophones tethered 
beneath the ice cover ["Eastern Arctic ambient noise", Pritchard, R.S. 
(1989)  In: Oceans '89, vol. 4, p. 1246-1251. IEEE Pub. no. 89CH2780-5].
Two omni-directional, calibrated hydrophones were deployed at 60 and 90 
meters beneath the pack ice at two sites roughly 1 km away from the ship 
and from each other.  A few ambient noise observations began 27 September, 
nearly continuous measurements began 10 October, and observations ended on 
18 November near 81 degrees N 32 minutes E when severe ice deformations 
made it impossible to maintain instrumentation cable continuity.  A 
PC-based LOFAR system measured the hydrophone output ["Real time ambient
noise spectra acquisition and display", Prada, K.E. (1986) In: 4th Working
Symposium on Oceanographic Data Systems. Proceedings, pp.199-207. Sponsored 
by IEEE Computer Society and Scripps Institution of Oceanography.]

2. Data Collection Methods

     Shielded instrumentation cable carried amplified hydrophone output into 
a Frequency Devices low-pass Butterworth filter with adjustable corner 
frequency (Model 901).  Data were filtered by an RC circuit to minimize 
signals below 1 Hz.  The acquisition system was a Compaq 386/20 computer on 
the ship.  Data were input through a Metrabyte A/D interface board (Model 
DAS-16G).  Computer peripherals included VGA monitor, 80387 numeric 
co-processor, 130 Mb hard disk and 135 Mb tape cartridge backup, and NEC 
color printer.

     The LOFAR computer program provided interactive control of data
acquisition.  Digitized input signals were sampled at the chosen maximum 
frequency.  The anti-aliasing filter corner frequency was set between half 
and two-thirds of the Nyquist frequency.  After collecting 1024 data points,
an FFT was performed, and the power spectral density estimated.  One hundred 
twenty-eight individual spectral densities were averaged, and the averages 
were recorded.  Narrow-band harmonics were removed, data were calibrated, 
and the resulting power spectral densities collected into one-third octave 
bands.  The first output bin was discarded from all spectra because of 
possible contamination by DC and other low frequency signals.

     For a frequency bandwidth of 512 Hz, data were input about 50% of time 
and FFT calculations required about 50% of time.  For these parameters, an 
average spectral estimate was obtained at roughly 2-minute intervals.  For 
all frequency bandwidths, the number of data points in the FFT was held 
constant at 1024.  The time required to input data decreased inversely with 
bandwidth.  The average spectral estimate was obtained at roughly 1-minute 
intervals at a bandwidth greater than 2000 Hz.  Since 1024 data values were 
used in each FFT, resolution decreased as bandwidth increased (for a 
maximum frequency of 512 Hz, frequency bins were 1 Hz, but for a frequency 
maximum of 8192 Hz, frequency bins were 16 Hz). 

     The system 'noise floor' was estimated by removing the hydrophone from 
the system and connecting all cables at the hydrophone location.  The system 
contained the shielded instrumentation cable, anti-aliasing filter, A/D 
interface board, and acquisition computer.  The noise floor was about 72 dB 
re 1 microPa**2/Hz across the spectrum, a threshold that remained rather 
steady throughout the experiment.  This noise floor probably resulted from 
the 12-bit A/D interface board, which limited the dynamic range to about 
39 dB.  All observed values are included here.  The user must eliminate 
those values contaminated by the noise floor.

     Data were recorded from only one hydrophone at a time.  The hydrophone 
indicated by H1 was tethered at 90 m depth.  Its location appears on the 
cover photograph presented by Pritchard and twenty-eight others [1990].  
The hydrophone indicated by H2 was tethered at 60 m depth.  It was located 
near the upper left corner of the same photograph.  The following table 
shows when data were recorded at each site. 
     
Hydrophone     Time when data began on hydrophone (UT)
     H1        10/08/88 1452 UT
     H2        10/20/88 2101
     H1        10/21/88 0817
     H2        10/21/88 0952
     H1        10/21/88 1139
     H2        10/21/88 1420
     H1        10/21/88 2208
     H2        10/22/88 0727
     H1        10/22/88 1008
     H2        10/22/88 1432
     H1        10/22/88 1900
     H2        10/22/88 2246
     H1        10/23/88 0718
     H2        10/23/88 1119
     H1        10/23/88 1614
     H2        10/24/00 0922
     H1        10/24/88 1421
     H2        10/24/88 2106
     H1        10/25/88 0903
     H2        10/25/88 1657
     H1        10/25/88 2038
     H2        10/26/88 0855
     H1        10/26/88 1435
     H2        10/26/88 1720
     H1        10/27/88 0655
     H2        10/27/88 1330
     H1        10/27/88 1947
     H2        10/28/88 0839
     H1        10/28/88 1310
     H2        10/28/88 2311
     H1        10/29/88 1718
     H2        10/30/88 0958
     H1        10/31/88 0411
     H2        11/01/88 1017
     H1        11/01/88 1706
     H2        11/02/88 2258
     None      11/03/88 1655  No hydrophone, noise floor tests
     H2        11/05/88 0213

3. Data Format

     Data are in the file AMBIENT.DAT in the directory \NOISE\AMBIENT on
the CD-ROM "CEAREX-1".  The data can be read from the archive using the 
following set of FORTRAN statements:

      PROGRAM ICINOISE
      DIMENSION FREQ(28), DATA(28)
      CHARACTER*80 FILENAME
      WRITE(*,*)' Enter name of ambient noise data file'
      READ(*,800) FILENAME
  800 FORMAT( A80)
      OPEN(UNIT=1,FILE=FILENAME,FORM='FORMATTED')
      READ(1,910)
      READ(1,910) (FREQ(I),I=1,28)
  910 FORMAT( 10X,28F10.2)
c     loop on time
 1000 CONTINUE
      READ(1,920,END=2000) TIME, (DATA(I),I=1,28)
  920 FORMAT( F10.4,28E10.4)
c
c     do something with the data here
c
c     end time loop
      GO TO 1000
 2000 CONTINUE
c
c     STOP
      END

     Noise intensity is presented for one-third octave bands from 2 Hz to 
1000 Hz.  Units are microPascal**2/Hz.  Average values over each one-third 
octave band have been divided by bandwidth and therefore describe spectrum 
level.  Missing values are indicated by data values of 1.E-9.  Time 
(Greenwich Mean Time) is presented in consecutive days of the year 
(decimal fractions), where 1 January 1988 at 0000 UT is TIME = 1.0000.

     Prior to day 272.5219, eight FFTs were averaged, giving roughly
7-second averages, in contrast to the 1-3 minute averages calculated later. 
The times of these closely-spaced data cannot be resolved by the resolution 
of 1.E-4 day, with the result that several consecutive records have 
identical times.  This limitation occurs only for the first 988 data 
points.

3. Results

     Ambient noise data were energetic at frequencies below 500 Hz at the 
outset, but the frequency band increased to over 5000 Hz as nearby ice 
ridged and rubbled.  Data at frequencies above 1000 Hz were contaminated by 
the noise floor, and have not been archived.  A visual comparison between 
the ambient noise and environmental variables showed that correlations 
exist.  Noise increased during high winds, deformations, and when stress 
levels increased in the ice.  A strong semi-diurnal tidal and/or inertial 
signal was also observed.  The concurrent wind, air temperature, current, 
large scale ice motion, local ice motion, ice stress, and under-ice noise 
data should provide opportunities to identify and describe noise generated 
by different processes.

     Two different types of analyses have been pursued:  statistical 
analysis of the data, and modeling.  Cousins [CEAREX Ambient Noise Data 
Measured Northeast of Svalbard. Naval Postgraduate School thesis, 1991] 
has estimated the 5, 50, and 95 percent noise levels, and the correlations 
between noise and local environmental variables.  Pritchard [Sea ice 
noise-generating processes. Journal of the Acoustical Society of America,
88(6), pp.2830-2842, 1990] has developed a model that accounts for local 
and distant noise sources generated by pressure and shear ridging, 
microcracking, and mixed layer shearing.

4. References

Cousins, J. D. (1991) CEAREX Ambient Noise Data Measured Northeast of
Svalbard.  Monterey, CA, Naval Postgraduate School, Thesis.

Prada, K. E. (1986) Real time ambient noise spectra acquisition and
display.  In: Fourth Working Symposium on Oceanographic Data Systems.
Proceedings, pp.199-207.  Sponsored by IEEE Computer Society and Scripps 
Institution of Oceanography, University of California, San Diego, CA.
IEEE Catalog No. 86CH2269-9.

Pritchard, R. S. (1990) Sea ice noise-generating processes. Journal of
the Acoustical Society of America, 88(6), pp. 2830-2842.

Pritchard, R. S. (1989) Eastern Arctic ambient noise.  In: Oceans '89,
vol.4, p. 1246-1251.  New York, IEEE. IEEE Pub. No.89CH2780-5.

Pritchard, R.S. and twenty-eight others. (1990) CEAREX Drift Experiment, 
EOS, Transactions of the American Geophysical Union, 71(40), (October 2), 
pp.1115-1118.

6. Contact Information

     Robert S. Pritchard
     IceCasting, Inc.
     11042 Sand Point Way NE
     Seattle, WA   98125-5846
     (206) 363-3394
     Telemail: R.PRITCHARD/Omnet

7. Acknowledgments

Acquisition of this ambient noise data set required the co-operation and 
help of many individuals, and I am grateful to all.  A.B. Baggeroer (MIT),
K.E. Prada and K. von der Heydt (WHOI) identified necessary hardware, and 
provided hydrophones and LOFAR software.  R.H. Bourke (NPS) provided the 
data acquisition computer.  T. Stanton (NPS) guided me in designing, 
assembling, and testing the data acquisition system.  M.D. Coon (BDM Corp.),
Chief Scientist of the first leg of the drift experiment, helped acquire 
and deploy the equipment.  A. Heiberg (UW) provided logistical support to 
the entire expedition.  J. Ardai and A. Gill provided on-ice help and 
expertise that eased the workload.  T.A. Jakobsen and crew of the 
POLARBJORN provided an efficient and pleasant ship that made it possible 
to concentrate on the scientific aspects of the expedition.  The entire 
scientific team worked diligently throughout the 10-week drift.  They 
showed remarkable resilience and good humor, even during breakup, when 
all our plans changed without notice.  This work was funded by the Office 
of Naval Research under contract N00014-88-C-0580 to IceCasting, Inc.  
I thank T.B. Curtin, Manager of the Arctic Sciences Program, for his 
encouragement and support. 

February 1991 (revised April 1991)