Geophysical Exploration of Crustal Structure

Exercise 6: Processing of real field data

Processing flow

1. Check headers
2. Band-pass filter
3. Amplitude recovery
4. Velocity (f-k) filter (Optional)
5. Deconvolution (Optional)
6. Statics Correction (Not mentioned)
7. Velocity analysis (Next exercise)
8. NMO correction and CMP stacking (Next exercise)
9. Migration (Next exercise)

Check dataset and headers

First, check the data headers of the data we are going to process. The actual data processing usually starts from editing the data headers. Fortunately, the header of the dataset here has already been finished editing. The headers contain the basic information of the dataset.

$ surange < Nshots.su

-----
19057 traces:                              : Number of traces
tracl    39069 58125 (39069 - 58125)       : Trace number
tracr    1 19057 (1 - 19057)               : Trace number
fldr     1687 2012 (1687 - 2012)           : Shot number
tracf    28 96 (96 - 28)                   : Receiver number
cdp      900 1300 (900 - 1300)             : CDP (CMP) number
cdpt     1 69 (1 - 1)
trid     1 2 (1 - 1)                       : Data type
offset   -2435 -170 (-170 - -2435)         : Offset distance (m)
scalel   -10000
scalco   -10000
counit   1
muts     0 11000 (0 - 0)
ns       5500                              : Number of samples in the trace
dt       2000                              : Sampling time (micro second)
day      206
hour     21 22 (21 - 22)
minute   0 59 (3 - 20)
sec      1 59 (45 - 3)
-----

Next, take a look at the data. Since the dataset is too large to display all, extract a portion of the dataset and display it.

For example, extract the traces whose shot number (in the header "fldr") is from 1800 to 1805, sort the data on the order of receiver number (in the header "tracf") and create a small subset of the data, such as "tmp_1800.su". The range of shot and receiver number and the name of the file are of your choice.

$ suwind key=fldr min=1800 max=1805 < Nshots.su > tmp.su
$ susort tracf < tmp.su > tmp_1800.su
$ rm tmp.su

$ surange < tmp_1800.su
$ sugethw key=fldr,tracf,cdp,offset < tmp_1800.su

$ suxwigb perc=98 < tmp_1800.su &
$ suximage perc=98 < tmp_1800.su &

• No surface waves are recorded.
• The waves near 0 seconds are the direct wave propagated in sea water.
• The waves below 6 seconds are the reflections from the sea bottom and the strata below.
• No reflected waves are observed between the time above.
• The water depth is about 4.5 (km) because 1.5 (km/s) x 6 (s) / 2 = 4.5 (km).
• Some traces are contaminated by noise.

cdp      1 2869 (1 - 2869).

Parameter estimation for preprocessing using test data

$ suwind key=cdp min=1095 max=1095 < Nshots.su > cdp_1095.su

$ suximage < cdp_1095.su perc=99 &
$ suxwigb < cdp_1095.su perc=99 &

Fig: Waveforms (enlarged) (Left) in an image, (Right) in wiggle traces.

Spectrum and filter

$ suspecfx < cdp_1095.su | suximage perc=98 &
$ suspecfx < cdp_1095.su | suxgraph &
$ suspecfx < cdp_1095.su | suop op=db | suxgraph &

• "suspecfx" calculates the amplitude spectrum via Fourier transform in time.
• "suop" applies various arithmetic calculations (operations) to the data. Here, the option "op=db" calculates a logarithm in dB.

Fig: Spectrum. (Left) The spectrum in a color image. The vertical axis is "frequency (Hz)". (Right) The spectrum in a graph style.

Now you can see the frequency components contained in the data.

Then, check the components in each frequency band. A simple shell script, create_BPF_panel.sh displays the frequency components at every 10 Hz, each has a 10 Hz bandwidth (0-10, 10-20, ...) Edit the filename in the script so that the script can use your data file.

$ sh create_BPF_panel.sh

Fig: Spectrum, (Left) Original, (Rights) After band-pass filtering of 0-10, 10-20 Hz, 20-30Hz, 30-40Hz, ......

Try applying a band-pass filter to the sample data. The example below uses 'f=10,20,70,80'.

$ sufilter f=10,20,70,80 amps=0,1,1,0 < cdp_1095.su | suspecfx | suximage perc=98 &
$ sufilter f=10,20,70,80 amps=0,1,1,0 < cdp_1095.su | suspecfx | suxgraph &
$ sufilter f=10,20,70,80 amps=0,1,1,0 < cdp_1095.su | suspecfx | suop op=db | suxgraph &
$ sufilter f=10,20,70,80 amps=0,1,1,0 < cdp_1095.su | suximage perc=99&

$ sufilter f=?,?,?,? amps=0,1,1,0 < cdp_1095.su > cdp_1095_flt.su

Fig: Spectrum in color display, (Left) the original, (Right) after band-pass filtering.

Fig: Spectrum in graph display, (Left) the original, (Right) after band-pass filtering.

Fig: Waveforms in image display, (Left) the original, (Right) after band-pass filtering.

Amplitude recovery

$ suximage perc=98 < cdp_1095f.su &
$ suop op=nop  < cdp_1095f.su | suattributes mode=amp | suop op=db |suxgraph &

• "suattributes" calculates various attributes of the waveform. 'mode=amp' shows the amplitude characteristics (so-called an envelope).
• "suop op=nop" does nothing (No OPeration) to the data. You can read the data in "suattributes" directly. This is used just for comparison with the next command line below.
• "suop op=db" calculates the logarithm in dB.

$ sugain tpow=2.0 < cdp_1095_flt.su | suattributes mode=amp | suop op=db | suxgraph &

• "sugain" applies a gain in amplitude.
• 'tpow=n' is a gain of n-th power of time (t). Here, 'tpow=2.0' indicates the gain is the square of the time.
• Gain is required in order to compensate for the effect of geometrical spreading, inelastic attenuation, and transmission loss.

$ sugain tpow=? < cdp_1095_flt.su > cdp_1095_flt_gain.su

Fig: Waveforms in amplitude display, (Left) Before the amplitude recovery (after the filter), (Left) after the amplitude recovery. Note that the amplitude after 6.0 seconds became flat.

Fig: Waveforms in image display, (Left) Before the amplitude recovery (after filter), (Left) after the amplitude recovery. Note the amplitude of deeper reflections increased.

Velocity (f-k) filter

The f-k spectrum is displayed for your reference.

$ suspecfk < cdp_1095_fg.su | suximage perc=98 &

Fig: f-k spectrum.

As the velocity filter will be applied to each shot gather (common-shot record), you will need to use a shell script to loop the filtering operation through all shot gathers.

Deconvolution

If you are interested in applying the deconvolution, visit here and follow the guidance. It is a good idea to compare the result with and without deconvolution.

This is the end of the parameter seeking for preprocessing. Keep the file you've created, for example, "cdp_1095_flt_gain.su" (if you applied the deconvolution, the file name should be "cdp_1095_flt_gain_decon.su" or whatever) since it will be used in the next exercise.

Execution of preprocessing

1. Apply band-pass filter,
2. Do amplitude recovery,
3. Apply deconvolution (optional),
4. Apply band-pass filter again (optional),
5. Resample the data to reduce the size of the dataset. Do down-sampling of the data from 2 ms of the sampling interval to 4 ms. You always need to apply an anti-alias filter before down-sampling to filter out the components higher than the Nyquist frequency. In this exercise, you can skip the anti-alias filter because you've already applied a band-pass filter.

$ sufilter f=?,?,?,? amps=0,1,1,0 < Nshots.su | sugain tpow=? | suresamp nt=2750 dt=0.004 > Nres.su

If you are to apply deconvolution, the command line should be something like below.

$ sufilter f=?,?,?,? amps=0,1,1,0 < Nshots.su | sugain tpow=? | supef minlag=???? | sufilter f=?,?,?,? amps=0,1,1,0 | suresamp nt=2750 dt=0.004 > Nres.su

Finally, sort the data on the order of 'cdp', then sort on the order of 'offset' in each 'cdp'.

$ susort cdp offset < Nres.su > Ncdps.su

-rw-r--r-- 1 watanabe staffs 214200680 Nov  7 12:17 Ncdps.su
-rw-r--r-- 1 watanabe staffs 423827680 Nov  7 12:17 Nshots.su