Geophysical Exploration of Crustal Structure

Exercise 4

NMO correction

This exercise uses the sample data "demo_nmo.su". Copy the sample data "demo_nmo.su" if you do not have one. Your reference is here.

First, display the data (a CMP gather).

$ suximage < demo_nmo.su perc=98 &

Apply normal move-out (NMO) correction with a constant velocity v=1500 m/s to this CMP gather, and display the result.

$ sunmo < demo_nmo.su vnmo=1500 | suximage perc=98 &

"sunmo" applies NMO correction.
Using the combination of 'vnmo' and 'tnmo', a 1-D velocity structure can be specified.
In this example, a constant velocity is applied because one velocity value of 'vnmo' is given.

Fig: NMO correction, (Left) the original CMP gather, (Right) After NMO correction with a constant velocity of 1500 m/s.

Investigate how the result of NMO correction changes according to the velocity by vnmo = 1600, 1700, 1800, 1900, 2000, 2100, 2200.

$ sunmo < demo_nmo.su vnmo=1600 | suximage perc=98 &
$ sunmo < demo_nmo.su vnmo=1700 | suximage perc=98 &
......

Too small velocity for NMO correction -> Over-correction (Right-up)
Proper velocity for NMO correction -> Proper correction (Horizontal)
Too large velocity for NMO correction -> Under-correction (Right-down)

Try a shell program create_CVS_panel.sh.html that performs Constant-Velocity-Scan (CVS). At the beginning of the script, you see,

------------------------------------
# Parameters
vmin=1200
vmax=2500
dv=50
------------------------------------

This script displays a series of NMO corrections using the velocity from 1200 m/s to 2500m/s with 50 m/s increments.

$ sh create_CVS_panel.sh

Fig: Image display of CVS panel. (Left) Original CMP gather, (Right) Result of NMO correction using the velocity from 1200 m/s to 2500m/s with the increment of 50 m/s.

NMO correction works well around the velocity = 1650, 1850, 2000, 2250 m/s. (It may not be easy to understand since the horizontal axis does not denote the velocity itself.) The stacking velocities are thus determined.

Try displaying the result in a movie.

$ suxmovie < cvspanel_demo.su n1=1601 n2=69 n3=26 loop=1 sleep=1 perc=98 &

"suxmovie" displays a series of images in a movie.
'n1=1601' denotes time sampling. SU can read this parameter from the header. So you can skip this.
'n2=69 n3=26' denotes that we have (64 data traces) + (5 empty traces for spacing) = 69 traces in the gather, with 26 velocity values.
'loop=2' loops forward and backward. 'loop=1' loops only forward direction. 'loop=0' never loops.
'sleep=' gives the time between frames in seconds.
Other options are similar to those of "suximage". Zoom is also available.

After NMO correction, stack (sum) traces in the CDP gather. Take a look at the effect of stacking.

$ susort d2 < cvspanel_demo.su | sustack key=d2 | suxwigb &
$ susort d2 < cvspanel_demo.su | sustack key=d2 | suxwigb key=d2 &

"susort" sorts the traces using header keywords. In this example, a header keyword 'd2' is used as the sort key. Here, the header keyword 'd2' carries the velocity value used in NMO correction. (See the shell script 'create_CVS_panel.sh')
"sustack" stacks (adds) the data. Traces that have the same value in the header designated by 'key= ' ('d2' in this case) are stacked together.

Under a proper velocity value for NMO correction, you can earn a large amplitude of the reflected wave after stacking the CMP gathers.

Fig: Waveform obtained by stacking after NMO correction. The horizontal axis corresponds to the velocity used for NMO correction.

Delete the file "cvspanel_demo.su" since the file is no more used.

Velocity analysis

In velocity analysis, a velocity structure is estimated using the NMO correction.

Display a semblance panel of Constant-Velocity-Stack (CVS).

$ suvelan fv=1200 dv=50 nv=30 < demo_nmo.su | suximage f2=1200 d2=50 &

"suvelan" creates a semblance panel of velocity analysis (CVS). The output is the trace amplitudes stacked using given velocities.
'fv=1200 dv=50 nv=30' calculates CVS with 30 velocity values from 1200 m/s with increment of 50 m/s.
Use the values of 'fv' and 'dv' of "suvelan" for the values of 'f2' (start value of horizontal axis) and 'd2' (interval of horizontal axis) of "suximage", respectively. Then, the horizontal axis can denote the velocity.
Middle-button-click displays the location (coordinates) where the mouse is on.
Pressing 's' key outputs the mouse location on a console. Do not click mouse buttons.

The stacked amplitude increases (blight color spot in the figure) where the NMO stacking velocity is correct. Then you can read the value of time and velocity by locating the mouse cursor on the red spot. Move the mouse cursor and press the middle-button.

Fig: Constant-Velocity-Scan (CVS) semblance panel. The vertical axis denotes 'time (s)' and the horizontal axis denotes 'velocity (m/s)' used for NMO stacking.

By designating a file name using the option 'mpicks=filename', the mouse coordinates are written in the file when the 's' key is pressed.

$ suvelan fv=1200 dv=50 nv=30 < demo_nmo.su | suximage f2=1200 d2=50 perc=98 mpicks=pick.txt &

!NOTE! Pick the points as 'time' increases (from top to bottom). The picked values will be used in the next step (NMO correction). You may sort the order of the picked values afterward as follows.

 
$ sort -n pick.txt > pick_sorted.txt

!CAUTION! Press the key 'q' on the keyboard to close the "ximage" window when you finish picking. DO NOT close the window by clicking close (X) buttun using a mouse or you will lose your picking result.

Then, change the format of the pick-file so that the picked values can be used as the input of the next step, NMO correction.

$ mkparfile < pick.txt > pick.par string1=tnmo string2=vnmo

Take a look at the contents of "pick.txt" and "pick.par".

$ cat pick.txt
$ cat pick.par

Options 'tnmo=1.0,1.5,2.0,2.5,..... vnmo=1400,1500,1600,1700,.....' describes the velocity structure. The velocity at time 'tnmo' is 'vnmo'. The number of the elements of the options should be identical.

NMO correction and CMP stacking

Next, apply NMO correction to the CMP gather. The velocity structure used for NMO correction is passed to the program as a pair of the arrays of 'tnmo' and 'vnmo' as below.

$ sunmo tnmo=1.0,1.5,2.0,2.5,..... vnmo=1400,1500,1600,1700,.....  < demo_nmo.su | .....

We have already created a file "pick.par" containing 'tnmo' and 'vnmo'. Here we use the parameters as below.

$ sunmo par=pick.par smute=5 < demo_nmo.su | suximage perc=98 &
$ sunmo par=pick.par smute=1.5 < demo_nmo.su | suximage perc=98 &

NMO correction cause a distortion (stretch) of waveforms where the offset distance is large (in case of 'smute=5'). 'smute=' option can remove such area. In the second case, waveforms stretched 1.5 times than the original were muted ('smute=1.5').

Fig: Result of NMO correction. You will see the reflected waves aligned horizontally.

Apply CDP stacking to improve S/N ratio.

$ sunmo par=pick.par < demo_nmo.su | sustack > trace.su
$ suxwigb perc=98 < trace.su &
$ suximage perc=98 < trace.su &

Stacking of the NMO corrected traces yields a single trace. Compare how the S/N ratio is improved from the original figure.

Fig: Result of NMO stacking. (Left) wiggle and (Right) image display of the waveform. In each figure, the result of stacking (5 traces) is shown on the left, and on the right is the original data. The figures show that the S/N ratio was improved.

A trace after CMP stacking is a trace that both the shot and the receiver are located at the CMP. The offset is zero. Therefore, this is called a "zero-offset trace (record)".

Homework

Create noisy data having a lower S/N ratio by adding more noise to the sample data "demo_nmo.su".

$ suaddnoise sn=5 < demo_nmo.su > demo_nmo_more_noise.su

"suaddnoise" adds noise to the data. Use any file name for the output.
The option 'sn= ' provides S/N ratio. The smaller value gives the lower S/N ratio (noisier). Choose any value for 'sn= ' as you like. A value between 5 and 10 seems appropriate. 'sn=10' is an easy way. 'sn<5' may be a tough choice.

Apply a velocity analysis to the data you created. Then apply NMO correction using the picked velocity values and apply CMP stack.

Submit

the image display of the CMP gather with noise,
and the waveform of your final trace (See tips below).
and the result of the velocity analysis (pick.txt or pick.par).

You will understand that the velocity analysis and CMP stacking are robust and effective even for the noisy data that the reflections are hard to distinguish.

Tips: How to display the stacked trace and the original trace in the same amplitude scale.

If you want to align the stacked trace (5 traces) and the original traces, concatenate the data using "cat" command.

$ cat trace.su trace.su trace.su trace.su trace.su demo_nmo_more_noise.su > stacked_result_and_original.su

Five stacked traces (trace.su) and the original CMP gather (demo_nmo_more_noise.su) will be displayed in one figure.

FYI

In the figures in this exercise web page, 5 null traces (zero traces) are inserted between the stacked traces and the original gather.

$ sunull ntr=5 nt=1601 > null.su
$ cat trace.su trace.su trace.su trace.su trace.su null.su demo_nmo_more_noise.su > stacked_result_and_original.su

"sunull" outputs null traces (traces of which the values are all zero).
The option 'ntr=5' indicates the number of traces. 'nt=1601' indicates the number of time samples in each trace.

You can find the number of samples using "surange" which outputs the summary of header values.

Exercise 3 < Exercise 4 > Exercise 5

Back

Last modified: Wed Oct 18 17:42:02 JST 2023