Rebecca was told to process some 1999 seismic data. She nor I where part of the data acquisition and there are no recorder's logs. She had several Exabyte tapes. Here's how "we" proceeded: 1) Ron Moe created a CD from one tape. He used dd, so the tape was in SEG-Y. Ron had to use a CD because Rebecca's computer doesn't have sftp.yr99jd85-0616z.segyTo download the SEG-Y file (17.7MB): Mac users: OPTION-click on the filename below (right button if multibutton mouse). PC users: SHIFT-left-click on the filename below.
2) What's in the file? >lsd seisdata.1 1 5 SHOT TR RP TR ID RANGE DELAY NSAMPS SI YR DAY HR MIN SEC 1 1 0 0 1 0 4000 6000 1000 99 85 6 16 48 1 2 0 0 1 0 4000 6000 1000 99 85 6 16 48 1 3 0 0 1 0 4000 6000 1000 99 85 6 16 48 2 1 0 0 1 0 4000 6000 1000 99 85 6 16 59 2 2 0 0 1 0 4000 6000 1000 99 85 6 16 59 Rename the file so it's more meaningful: >mv seisdata.1 yr99jd85-0616z.segy 3) The plot (script) has a space between every shot (3 traces) and uses a generic .5 sec agc window. 4) Adjust the parameters: a) Use stime 6 and nsecs 2 to plot only the data of interest. b) Use ann shotno taginc 3 to annotate the shot number on every third trace. 5) Adding N traces together will reduce the random noise by SQRT(N). While this reduces the horizontal resolution, I suspect the trace spacing is small compared to the shot spacing. A mix might not be good on data with steeply dipping events. Fake sioseis into thinking the data are cmp gathers rather than shot gathers by using parameter ntrgat 3 in process diskin. Sum the traces within the gather by using process stack. Change the plot parameters for displaying continuous subsurface (remove fspace, spacei, nspace and change taginc). New script. 6) I suspect these data were collected using GI guns as a source, so try a filter passband of 20 250. Adjust the plot horizontal scale (parameter trpin) and reduce the trace amplitude (def). Get rid of plotting the trace wiggle (wiggle 0). Overdrive and clip the amplitudes (make def too big while limiting the deflection by clipping it). Increase the agc window. New script. 7) Too much clipping. Decrease def and get rid of clip. Also note the trace blanking just prior to the water bottom, which is caused by agc trying to make the noise in the water column the same size as the water bottom. Use agc parameter center .1 to minimize that. Adding a 3 trace mix will further reduce random noise by SQRT(3). Mix is a running average of adjacent traces. It good when there's not much dip but will "smear" whe dip is present. Script for new plot. 8) On a different seismic line (over a seamount), the discussion went: This mornings parameter changes for seisdata.2 are in plot. Use stime 0 and nsecs 8. explanation: sioseis plots a 3(?) trace gap whenever the deep water delay changes (the time of the first sample in the trace - why record the water column?). By using stime 0, sioseis forces every trace to the actual time by adding zeroes. 9) Time to step back from the processing and think about the geophysics and geology. The extremely steep walls of the seamount will not have sediment nor will there be any reflections because the seismic energy wavepath is past the critical angle. The litle hyberbola (or frowns as we call them) are caused by point sources - diffractors - faults and other discontinuities. Also, remember that seismics are omni-directional; what comes in first is not necessarily what's directly under the ship. It might be off to the side or in front or behind. I filtered with pass 40 150 and noticed diffractors in the middle of the seamount! Hyberbolas can be collapsed with migration, but we'll need to know the shot spacing. The higher frequency of the diffractors indicates that these are from the surface, so an fk migration with water velocity might work great. The next step is to pick the water bottom with program sioplt so we can mute (zero) the water column and make prettier pictures for publication.