The problem
Determining which traces are bad
Screenshots:

---

A tar.gz archive is available which contains a small example of all of the steps outlined below. It should make this process more clear »upload/trinterp.tar.gz. This job will only work with Claritas, a similar program is available for Seismic Unix called suinterp.

The problem

• The DOBS instruments will record seismic data and store in memory until it reaches 2 megabytes in size. At this point it will write this data to hard disk. The instrument will stop recording while writing to disk, which results in either dead traces (traces with data values of 0) or missing traces.

Determining which traces are bad

• Claritas has a process called automute which is able to automatically mute bad traces, but in this case we're going to do it manually with the SQC tool.

• Read in the segy file (or other) and write out as a claritas disk file using readsegy, missing, and discwrite (these are contained in trinterp_part1.job). The missing process inserts dummy traces so the RECORDNUM header fields increase monotonically. We will use this output file as input to the sqc program by typing the following command at the terminal.

sqc sqc_file.clr

• Fill in a couple of parameters such as output file (dead_traces), and number of trace to display or maximum ensemble (200 or less). Click on the kill button, and specify RECORDNUM as the secondary key. Recordnum corresponds to the shot number. Start to select traces to kill by clicking on them. Right click to stop. Be sure to 'kill' all dead traces otherwise some dead traces might stay around during later stages (holding the left mouse button down will allow you to select multiple traces).

• When complete, exit and the .tre file will be output.

• Run trinterp_part2.job to fill in the killed traces with interpolated data.

• Run trinterp_part3.job to see the difference between the original data and the data with interpolated traces. The original data will be displayed on panel 1, and the new data on panel 2.

NOTE: the traces inserted as part of the missing process do not contain any header information for offset, therefore operations such as a reduction velocity or nmo will not work properly on these traces.

Screenshots:

• Dead traces are apparent in the original data.

• The dead traces have been selected in sqc, and killed with tredit.

• The killed traces have been filled in with interpolated traces (along with a forward and inverse nmo static correction).