GPH 492/692 - Seismic Reflection Processing Lab

John Louie, September 1, 2011
Copyright © 2011 John N. Louie, all rights reserved.

Key Skills

When you complete this lab exercise, you should have gained experience toward the following skills.


120 points total
  1. (10 points) Load a 3 Mb seismic reflection data volume into Viewmat.

    A geophysics senior at Victoria University of Wellington, New Zealand recorded this high-resolution survey along a beach in the Seatoun suburb of Wellington. Seatoun sits in a small fault-bounded basin, and has recorded the highest earthquake-shaking amplifications of anywhere in the city. This survey complemented previous gravity work to confirm basin geometry and examine the stratigraphy for evidence of recent faulting.

    Try the Animate button, or drag the Plane Index slider at the top of the plot to see all the records. Answer this question: what type of wave has the most energy? This energy is stronger in the Seatoun records than in the Dixie Valley data you saw before because in Dixie Valley we used blasts in 3-m-deep holes, while in Seatoun the impact source was on the surface.

  2. A Geometrics R-24 system stacked 10 sledgehammer hits per source point from 24 channels of single 12-Hz geophones spaced at 8 m. Offsets and the line progression were from west to east.
    1. The records are 1 sec long; your computer may have an easier time if you just take 0 to 0.5 seconds of time. Select Methods->On Each Plane->cutTime and specify that Traces Start at 0 and Trace Length 0.5 sec, and press the cutTime button. After the new record display pops up you can select, on the old display, File->Close Window to save memory. Also select Edit->Clip at RMS on the new display.

  3. Try an automatic gain control to see the reflections more clearly. Select Methods->On Each Vector->agc and enter a Window Length of 100 samples. It's helpful to know that since the sample interval is 1 ms, the traces are 500 samples long. Press the agc button; then select Edit->Clip at 3*RMS on the data window. Note that the slow waves have decreased amplitude but the noise before the first arrival is now equally strong. Animate through the records again.

  4. For filter testing, you will want to select Methods->In Place to uncheck and turn off the substitution of a method's result over the original data. The next time you look at the Methods menu In Place should not be checked.

    Use the Methods->On Each Vector->bpfilter selection to make filter tests. For each test frequency band, leave Pad at 2 (takes longer but avoids ``wraparound''; you can see what happens with Pad=1), and verify that Dt is correct at 0.001 sec. The next 4 fields specify a trapezoidal frequency response. Nothing will be left below the frequency specified in the first field (Lowdown); response will ramp up linearly to the frequency in the second field (Lowup); 100% response will be maintained between the frequencies of the second and third (Highup) fields; and response will ramp down to nothing at the frequency of the fourth field (Highdown).

    Each ramp at the ends of the pass band should have a ``20% taper''. So, if I wanted to pass the 50-100 Hz band, I would enter 40, 50, 100, and 120 Hz into the four fields, respectively. At 0 Hz and the Nyquist frequencies (the defaults shown when the bpfilter dialog pops up) you do not need any tapering.

    1. Try the 50-100 Hz bandpass filter with 20% taper suggested above. Since the filter has removed amplitude, there is less on the new display that pops up. Re-clip it at 3*RMS.
    2. Try the same filter with hardly any taper by entering 49, 50, 100, and 101 Hz into the 4 fields. If you compare, say, the two filters on Record 1010 at Plane Index 7, you will see that the reflection is less sharp, more ``ringy'' if you do not use a 20% taper. This is known as Gibbs's Phenomenon - if you try to filter too sharply in frequency, you end up with more energy at frequencies right where you were trying to cut them out. Close the window with the untapered filter results.
    3. Make yourself some notes of what phases you can and can't see in your 50-100 HZ tapered filter result. Now you can close that window. Going back to the original tegain'ed data display, use bpfilter to make several filter tests. Keep notes on each result so you can close the filtered data window before you try another one, to avoid memory problems. I suggest trying bands of 0-25, 25-50, 50-75, 75-100, 100-150, 150-200, 200-300, and 300-500 Hz. Use a 20% taper. Note: AC power in New Zealand cycles at 50 Hz. Since this survey was in an urban area, the seismic cables pick up that frequency inductively.
    4. (10 points) Examine the bandpass-filtered test records for the frequency range of the refractions and reflections. What frequencies show the reflections best? Are there any other types of waves that share this frequency range? Assuming that the reflections arise at depths where Vp = 1800 m/s, what is the range of reflection wavelengths implied by their range of frequencies?
    5. (10 points) Decide on a frequency band that will show the reflections best while mitigating other waves and noise. Use a taper of 20%, or more if that is helpful. Make a final filtered set of records, save it as a JRG Pack, and then close your original records. You will have to re-apply tegain and also re-clip the display. Turn in one nice example record (by email); use File->Save Images->Write PNG File (name the file Team?-Q4e.png), and make sure you write down the 4 corner frequencies in Hz and turn them in too.

    Useful equations:

V sub a ~=~ {DELTA x} over {DELTA t} ~~~~~~
V ~=~ f lambda ~~~~~~
DELTA g sub max = {V sub min} over {2 f sub max} ~~~~~~
V sub R ~=~ 0.9 V sub s ~~~~~~
V sub p ~approx~ sqrt 3 V sub s
DELTA g sub max ~=~ {V sub a} over {2 f sub max} ~~~~~~
sin theta ~=~ V over {V sub a} ~~~~~~
DELTA z ~=~ lambda over 4 ~~~~~~
DELTA x sub f ~=~ V over 2 sqrt {{t sub 0} over f}

  5. (20 points) What is the range of apparent velocity shown by the reflections? For the minimum reflection apparent velocity and maximum reflection frequency you find, compute the maximum Deltag needed to avoid spatially aliasing the reflections. At this minimum apparent velocity, assuming a rock velocity at the surface of Vp = 1000 m/s, at what angle from the vertical theta are the reflected waves hitting the receivers?

  6. (20 points) On a PNG or JPG image of a bandpass-filtered test record, mark the shallowest and the deepest reflections you think you can identify (or describe the record numbers, offset, and time of a couple of examples). What approximately are the two-way travel times t0 of these reflections at zero offset? What are their dominant frequencies? For each of the two reflections compute its vertical resolution Deltaz using the Widess criterion, and its Fresnel radius Deltaxf. Assume V = 1800 m/s. Would the upper or the lower reflection locate potential faults more accurately?

  7. Estimate stacking velocities for NMO correction from sixteen constant-velocity stacks between 1000 and 2500 m/s.
    1. Create the suite of 16 CV stacks from your best filtered set of records. Make sure you have fixed their coordinates. Select Methods->On Each Plane->cvstack. Note whether the correct offset range appears. Scroll through the text box at the top to read about the parameters required. Since the R-24 and the SEG-Y file in this case only provides source and receiver X-coordinates, appearing to be a purely east-directed line on a map, the cvstack dialog suggests the midpoint line run at N90E from the ``westernmost'' shot or receiver point. You will have to change more of the dialog's suggestions for this data set, like setting nm=37, dm=8, and arc=16. In addition: 1) make sure v0 is the correct minimum of your desired velocity range - so change it to 1000 m/s; 2) change nv to 16 because you want to test that many velocity values; and 3) change dv to 100 m/s, so v0 + (nv-1)*dv will now give the correct 2500 m/s velocity maximum you want to test.
    2. Click the cvstack button. As the computation proceeds, you can follow its progress from the output in the DOS window or Java Console. CV stacking can take several minutes.
    3. You can save and later restore your CV stack volume as another JRG Pack. Later, when you want to work on the stacks further, start Viewmat and select Load JRG Pack, and you can proceed with picking.
    4. With the volume of CV stacks, identify the stronger reflections. Where the reflections are strong you should be able to identify several separate reflections in a small area. Use the Animate button, or the Plane Index slider to page through the stacks at different velocities, focusing on small areas.
    5. You may notice there are few reflections stacking in below 0.3 sec. Use the cutTime method now on the CV stacks, and increase the plot's vertical exaggeration to 0.1, to get a nicer display.
    6. (10 points) Turn in a copy of one of the CV stacks indicating the reflections you will pick.
    7. Now pick the stacking velocity of each of your selected reflections by flipping through the suite of CV stacks, concentrating on just one reflection at a time. Pick the distance, time, and velocity where each reflection is strongest and most continuous. Your picks appear in the same pick window, but these are NMO velocity picks, not first-arrival picks. (So exporting for SeisOpt would not have any meaning.) You can still, of course, copy and paste or Save your pick text to a spreadsheet. Just remember that you have to press the ``Show All'' button at the bottom of the Pick Window to see all the picks from all the planes (all the trial velocities) in one list. The Save button does write all the picks to a text file, which you can then open in a spreadsheet or Notepad. Saving a JRG Pack also saves all your picks.
    8. The pick text has the following column order: Amplitude; time index; time in seconds; trace index; trace distance from VP0 in meters; plane index; and NMO velocity in m/s; followed by location information for the midpoint.

  8. (10 points) Check your velocity picks by computing interval velocities for the intervals between picks. From the filtered data records window (not the CVstack), select On Each Plane->makevels. Select ``Dix Interval Velocities'' instead of the default ``Interpolated RMS Velocities''. Use ``nv=1'', and you will get just one velocity section plane- you don't need any Vnmo Adjustment. Examine the interval velocity section for correctness. The velocity section that comes out is color-coded for the velocity value. It should appear blocky- if it's smooth then you must have selected ``Interpolated RMS Velocities''. Scroll down to look at the color bar at the bottom, and note the highest velocity. You can make picks on the velocity section and see the velocities in the first column of pick text. Interval velocities above 5 km/s and negative velocities are unacceptable, but can easily result from reasonable picks. The picks are too close together in time to account for such a higher velocity on the lower pick. Alternatively, if the picked velocity decreases too fast, the Dix equation can go negative under the radical, and the section will have black area of zero velocity. The black areas of invalid Dix computation can be thin horizontal lines; you can use the On Whole Dataset->clip method on the velocity section to make sure the minimum velocity output is above zero (then Cancel the clip). Turn in a Dix velocity section with problem areas identified.

    Useful equations:

V sub i sup 2 ~=~ {V sub {i^rms} sup 2 t sub i ~-~
V sub {(i-1) rms} sup 2 t sub {i-1}} over {t sub i ~-~ t sub {i-1}} 
V sub rms sup 2 ~=~ {sum from i=1 to n V sub i sup 2 DELTA t sub i}
over {sum from i=1 to n DELTA t sub i} ~~~~~~
DELTA t sub i ~=~ {h sub i} over {V sub i} ~~~~~~
DELTA t sub i ~=~ 1 over 2 ( t sub i ~-~ t sub i-1 )

  9. (10 points) Now adjust your stacking-velocity picks so the interval velocities will come out more reasonably. (If you had gotten reasonable interval velocities the first time around, explain how much your picks could vary yet still produce reasonable interval velocities. You got good velocities, but what is their error?) Try making the velocity difference between nearby picks in the same column as small as you can possibly justify given the reflection images in the CV stacks. Change your velocity values in notepad; you don't have to change the plane index values. Just make sure the pick text has the original column order. You don't have to paste in the columns after the velocity column. The makevels and cmpstack dialogs only look at the first 7 columns, and only reads the time, distance, and velocity columns. Just make sure you have been working with all of the pick text, from all the CV-stack velocities. Go back to the filtered records and plug the adjusted picks into Methods->On Each Plane->makevels. If these velocities are still unreasonable, re-adjust your picks and try again. Turn in your picks and computed interval velocity section.

  10. (20 points) Now use Methods->On Each Plane->cmpstack to compute a final stacked section from your filtered shot records (not from the CV stacks). The main difference from the cvstack dialog is that instead of specifying a range of constant velocities you copy and paste in your final, adjusted velocity pick text. Check the default values for the other parameters.

    Using your final stacking velocity at the largest time you are displaying in the stack (say, 0.3 sec), adjust the plot's vertical exaggeration so it will have approximately no VE after accounting for this time-to-depth conversion. Label the plot with depth instead of time, and turn in a plot of this approximate depth-converted final stack.