DESCRIPTION OF THE WINNOWING PROCESS

The original 1/10-second ascii data files were winnowed to
produce 1-second data.

The winnowing process rejected 9 of every 10 data points starting
with the 2nd point of each flight line.  Stated inversely, the
1st, 11th, 21st, and so on data points of each flight line were
selected to produce the winnowed data files.

The winnowing process effectively lowered the nyquist frequency
of the data spectrum from 5 Hz down to 0.5 Hz.  Therefore, a
despiker and low-pass filter had to be applied to the altimetry
and magnetics data channels before the winnowing process could
occur.  The filter reduced frequencies above the 0.5-Hz nyquist
to insignificant amplitudes so that aliasing could not occur.

Of course, some detailed information was lost in the filtering
and winnowing process.  Therefore, if any profiles are to be
re-processed or examined in minute detail, the original
1/10-second data should be uncompressed.  However, keep in mind
that, since the survey was flown at 1000 feet above ground level,
the anomaly "half-width" of a surficial dipolar source would be
about 1.5 seconds.  That is to say, the 1-second sample rate of
the winnowed data can easily define the basic shape of any real
anomaly.  The 1/10-second data is only useful if a higher order
fit of the shortest anomalies is required.