Clark and Yarnall 



To avoid confusion Table II has been prepared which specifies the 

 intervals of data collection, designates the geophysical categories, 

 i.e. , the headings under which data from a given interval will be 

 discussed, and indicates previous publications which deal with some 

 of these test intervals. 



An additional source of confusion may be the wide range of 

 time scales that must be employed to adequately illustrate phenomena 

 with characteristic periods in the range of seconds (surface waves) to 

 the range of weeks and months (atmospheric phenomena) . The major inter- 

 val of data collection, LCT-1, covers almost 5 months, and represents 

 the longest time base on which data is to be illustrated. Certain of 

 the data from the long test have already been introduced in Figure 13. 



Signal phase at H3 and H43 were each referenced to an arbi- 

 trary starting point at the start of this long test in September. 

 The arbitrary references were re-established after a 5-day inter- 

 ruption due to Hurricane Inez. In regard to the H43 phase curve, 

 three time scales of events are resolved in this figure: the long 

 down trend of over 250 cycles extending the full length of the test, 

 the prominent peaks in the curve with characteristic periods on the 

 order of days (speculatively attributed to atmospheric phenomena of 

 the same characteristic period), and the "high frequency" fluctuation 

 which is due to the tidally related phenomena. 



On the time scale of Figure 13 it is apparent that the tidal 

 effects have been reduced almost to the status of high frequency 

 "noise" riding on the lower frequency spectral components of the 

 curve. To more effectively illustrate the tidal effect, two standard 

 techniques have been employed: selected portions of the data have been 

 presented on an expanded time scale; high pass filtering has been 

 employed to remove the dominant low frequency spectral components. 



A further expansion of time scale is necessary to effectively 

 illustrate the signatures of short period internal waves, which occur 

 most frequently with characteristic periods in the range of a few 

 minutes. The surface wave signatures, with characteristic periods 

 in the range of seconds, appear as only an unresolved smear in all 

 figures which illustrate lower frequency events. One additional ex- 

 pansion of time scale is necessary to illustrate these signatures. 



Whenever practical, digital techniques have been employed 

 for spectral analyses of the acoustic phase and amplitude fluctua- 

 tions. It has been convenient to compute power spectral estimates 

 based on the method devised by Evans, ^ al,^^ which corrects for 

 limited forms of non-stationarity in the data. In the study of the 

 tidally related effect, sharp spectral lines are involved and a power 

 spectral estimate may not be the most efficient data analytical 

 approach. The application of an improved technique to these spectral 



320 



