Notice that channel 21 is funny.  It looks like the polarity is reversed.


Step 3: Use process weight kill the "dead" traces.  I used nmo so that
it's easier to track an "event".  Polarity inversions are much easier
to see if the first break is lined up.  I used the same scalar on each
plot so that any gross change in shot amplitudes would be obvious.
The trace 21 reverse polarity was not on any of the shots on tape 39.

Step 4:  Look at a single channel plot of the first 2000
shots using channel 2.  Note that the plot is annotated by GMT
and I used process GAINS rather than AGC.

Step 5:  Sort out the geometry/navigation.  ASA provided a smoothed
navigation file.  The onboard GPS was a Y-code (military) receiver, so
the navigation is probably pretty good.  The SIOSEIS geometry option
type 6 uses a navigation file to locate each shot based on the GMT
time in the SEG-Y trace header and the GMT in the navigation file.
I have to assume that the GPS time was used by the OYO seismic recorder.
I wrote a PERL script to convert the Palmer navigation file to a SIOSEIS
navigation file.

Step 6:  Examine the frequency spectrum to see if the data can be
decimated.  This required a rewrite of the Matlab progarm rdsegy.m
used in another example because the machine being used had Matlab5.1
and rdsegy.m didn't work.

Step 7:  Read all the data to disk, file at a time, decimate the date,
inventory each decimated file, plot the "short" trace in groups of
2000 shots, and finally store the decimated file on the IGPP mass store
(e.g. rcp tape40 arch:/archive/mcs/1997/nbp9702).

Step 8:  Determine what data should be stacked from the "short" trace
sections made in step 7.  Gather the data into rps.  Since I was on
disk space, I decided to throw out the "bad" traces.  Process weight
sets the SEG-Y dead trace flag when a weight of zero is used and
process gather drops dead traces from the gather.  I also decide to
flip the reverse polarity trace at the same time so I wouldn't have
to worry about it later.

*******    CAUTION    *********
SIOSEIS creates temporary files that can be destroyed by running
multiple SIOSEIS jobs in the same directory simultaneously.
Do not run multiple job simultaneously in the working directory.


Step 9:  Velocity analysis.  We need to determine which velocity
analysis method is best for this dataset.  The first one is semblance
contoured by Matlab.  After modifying the vpick script to work on
only three seconds of data and not to filter the data, I ran it with
just one rp just to make sure it was working correctly.

This analysis showed that the bad traces I picked earlier from tape 37
are not applicable to tape 39.  Trace 21 is not reversed either.  Change
the weights and regather.

Looking at stdout (standard output is the printed output), there's an
error message:
  SEG-Y Shot time error on shot  2930 time  68   1   17   33   242.0000
The lsd done earlier show 4 minutes of data missing.

   While picking the semblance contours, I could see hints of some high
velocities for basement.  I may want to repick, or try a different
velocity analysis technique.

   The vpick output filename was:   gathers.day068-0000z.vpick
and was the cut and pasted into plot script.

Step 10:  Check the nmo velocity picks by applying the velocity
functions to rps they were picked from.  Note how the same script is
used to plot every 100th rp, except this has nmo agc added in.  I added
agc so that I could see any events.  The velocities had to be edited
because I hit the mouse twice in the same spot once, but the sioseis
edit caught the error.

Step 11:  Muting the NMO strect can be done by the amount of moveout
or by specifying the mute pattern.  Specifying the mute pattern 
becomes difficult when the water bottom is constantly changing, so I
decide to try the automatic water bottom picker (process wbt), however
this becomes hard in noisey data.  I used the process prout to print
the amplitudes of rp 1000 tr 1 to see what the amplitudes were.  The
water bottom is very strong in comparison to the data immediately
after it, thus the traditional long window approach doesn't work.

     The variation of picks is troubling, so I need to try some other
method.  The dump of the amplitudes shows that the water bottom 
reflection has an amplitude consistently above .1 while the background
is less than .1.  

      Several traces were picked early, including rp 1011 trace 1.

Step 12:  Determine a good mute pattern by plotting some moved out
data with an agc, so we can see all events clearly on a single shot.
Annotate every trace by it's range.

Step 13:  Check all the prestack parameters on every 100th rp.
There are some minor things that could be done, but I'm anxious to see
the stack.  One rp didn't have the water bottom picked correctly. 
There was a trace with lots of low frequency (streamer strum?).  A
couple seemed to be have basement undercorrected.
chkstk gathers.day068-0000z 1000 100 2900

Step 14:  Stack the data and plot it so we can see if we're making
any progress.

Step 15:  Looking good!  The next thing to try is to get rid of the
noise; the 10Hz stream strum probably can be filtered out by using
a higher pass on the low end (we did a 10x80 filter during decimation).
We also need to get the plot of the stack data at the same horizontal
scale as the short trace plot (40 traces per inch) because the sort
by rp created 4 reflection points for each shot.
     In order to change the horizontal scale of the plot, I had to
change the plotter specified by parameter nibs.  The early plots
used nibs 75 or 75 dots per inch so that the plot file was small.
My favorite 300 dot plotter is the HP DesignJet, nibs 2848.

Step 16:  Try a "median stack", which uses the median value of all
traces at each time sample.  I guess this data is too noisy, because
the regular stack with "despiking" is better.  The problem with
noise is that migration will make "smiles" out of each spike in
the water column.

Step 17:  Pick velocities every 100 rps for data from day 068 0000z
to 1500z.  I made a separate disk file for groups of 50 rps
(e.g. rps 5000 to 10000 with an increment of 100) to keep the disk
access time down.  Picking 50 velocities took me 30-45 minutes, so
I wanted a break about then anyway.  Notice that I filtered the data
with a narrower bandpass filter during the extraction process.

Step 18:  Stack the data from 0000z to 1530z, but display it in 2 hour
chunks.  The new horizontal scale is 300 dots per inch because the
plotter is 300 dots per inch.  Anything higher than 300 dpi will cause
some data to be lost.

 
Step 23:  Convert three SIO rasterfiles to a single HP-RTL file and
send to the HP plotter.

Step 24:  Create a seismic plot with the magnetics above the seismics.
The plot must be black and white and should be small enough to publish.
    The most compressed horizontal scale is to make each plotter line
contain a seismc trace.  The HP DesignJet has 300 dots per inch, so
the most that can be done is 300 traces per inch (trpin 300).  I 
decimated the data by retaining every fourth trace.
The start time is set to zero (stime 0) so that the magnetics can
be centered around the 1 second time line.  The three migrated lines
are files fdm.line5x4,   fdm.line6.1x4,   fdm.line6.2x4.
The three SIOSEIS rasterfiles are: siofil5x4, siofil6.1x4, siofil6.2x4

    Two of the three lines are stored in reverse order so the east end
of the seismic line is the first trace to be plotted.  This creates a
problem when associating the magnetics with the seismics since the
magnetics are ordered with time increasing.  e.g.  magnetics:
1997     47      0       1       0        -116.0
1997     47      0       2       0        -113.0
1997     47      0       3       0        -111.0
1997     47      0       4       0        -116.0
1997     47      0       5       0        -118.0
1997     47      0       6       0        -109.0
    The magnetics are all in one large file, so I made a smaller file
corresponding to the times of each seismic file. The three maggie 
files are: maggie.line5, maggie.line6.1, maggie.line6.2

     I also made a listing file for each SEG-Y migrated file.
The three lsd file are: lsd.line5x4,  lsd.line6.1x4,  lsd.line6.2x4

    Program overlay plots "picks" relative to the first trace.  Each
successive "pick" is located by the number of traces away from the
first trace.  Overlay uses trpin (traces per inch) and the plotter
resolution, dots per inch, to compute where the pick is by plotter
lines.  I wrote two PERL scripts to merge the magnetics file and
the seismic listing file to create a file with the parameters for
program overlay.