When wave spectra derived from both gravimeters are plotted on the 

 same set of coordinates for direct comparison, corrections for instrument 

 response must be applied. This was done by computing the (acceleration)^ 

 response of each instrument and then applying this directly to the spectra 

 plots. A similar correction derived from the microbarograph response curve 

 (figure 3), although small, was applied to the micropressure spectra. 



d. Length of Time Series 



The nijmber of 5-second observations used for the spectral analysis 

 of each time series was a compromise between the best spectral estimate 

 possible and the length of time that the power spectnim remained stationary. 

 As mentioned above, the greater the number of observations for a given number 

 of lags, the greater the number of degrees of freedom or the closer the esti- 

 mated power spectrum is to the true but unknown spectral density. Examination 

 of individual records and their power spectra show that for intervals greater 

 than 45 to 60 minutes the spectra vary somewhat with slight shifts of peak 

 energy with time. This can be seen when any of the 90-minute wave or micro- 

 pressure records are divided into two 45-minute sections. Spectra for the 

 two consecutive 45-minute time series of each record are compared in figvires 

 ^ and 6. 



A 45-minute record appears to be optimum for obtaining stationarity 

 and sufficient data points for a meaningful analysis. This was tested by 

 dividing a typical 90-minute record into four equal, consecutive time series. 

 Comparison of the frequency distribution of observed displacements for each 

 adjacent series shows a good degree of stationarity from one series to 

 another. Thus, 45-minute series have been used when the spectral distribution 

 is more important than the overall estimate of spectral energy. For the 

 latter, 90-minute records are used; this essentially doubles the number of 

 degrees of freedom. 



4. THE 2-ELEMEMT ARRAY 



a. General Description 



The LaCoste-Romberg gravimeter #22 was installed at the ARLIS II 

 campsite, and automatic recording began on 31 May 19^1 . On 3 June I961 the 

 World-Wide gravimeter #5 was set up on an ice floe adjacent to the ice island. 

 This station was located l,2i|0m from the campsite on an azimuth of l43°T. 

 The World-Wide gravimeter was read visually every 5 seconds for 90 minutes 

 beginning 021 5Z. A portion of these two records is shown in figure T» An 

 attempt was made to match amplitude peaks of the same phase. Althovigh the 

 choice was quite tentative, it suggested that a given phase arrived at the 

 ARLIS II station 30 to 35 seconds before it did at the pack ice station. This 

 would indicate an apparent average phase velocity of 33 meters per second. 



The 90-minute record taken with this array was divided in two, and the 

 spectra of each half were computed (figvires 8a and 9a) • The coherences and 

 phase angles of the two sets of records were computed as described above and 



17 



