3-D VSP Survey Design and Processing
3-D VSP Survey Design and Processing
3-D VSP Survey Design and Processing
ABSTRACT
This paper studies the design of 3-D VSP surveys. It also outlines a processing
procedure to create 3-D images. Two numerical models (a dipping layer and a dome
model) are used to test the effects of variable shot and receiver positions. We find that a
high fold can be achieved near the well for realistic shot and receiver distributions. In
addition, we evaluate an synthetic data the 3-D VSP coverage of a current 3C-3D
seismic survey. A 3-D processing flow which includes NMO correction, muting,
binning and stack makes excellent images of the targets.
INTRODUCTION
There is often a need for a high-resolution, three-dimensional (3-D) seismic picture
around the borehole. Both 3-D VSP (using an areal distribution of surface sources and
a downhole tool ) and 3-D reverse VSP (with a surface array of receivers and downhole
source) hold great promise for near-well imaging.
In situ seismic measurements (VSP, reverse VSP and crosswell surveys) have
proven useful for imaging and estimating rock properties near or between wells.
However, because the earth is generally heterogeneous in three dimensions and our
regions of interest are usually volumetric, 3-D images are critical. Economical surveys
are also essential. Thus we must pursue practical measurements which provide 3-D
images in a cost-effective way with minimal production impact. We would argue that
the 3-D VSP is the current best candidate to solve this problem. Furthermore, the 3-D
VSP can piggyback on a conventional 3-D surface seismic survey by passive and
simultaneous monitoring of the surface shots. The reverse VSP could also be co-
recorded into the 3-D surface geophones, if there were an acceptable downhole source
and if its shooting didn't slow the surface survey down too much. There is
considerable practice of and literature on 2-D VSP surveying. In addition, a number of
authors have recently considered the analysis of 2-D RVSP data (Jackson et al., 1989;
Jones, 1991; Kragh et al., 1991; Naville et al., 1991; Hardage, 1992; Parra and Bangs,
1992). Haldorsen et al. (1992) used several 2-D surface lines to image data recorded
from a downhole drill-bit source. Their drill-bit seismic image matched surface seismic
and conventional walkaway VSP images reasonably well. Aleotti et al. (1994) also
showed drill-bit images (prediction ahead of the bit seismograms) that were similar to
synthetic seismograms produced from logs after the well had been drilled through the
predicted depths. Less has been written on 3-D VSP. Chen and McMechan (1992)
considered the 3-D case and used synthetic data from a salt structure model to perform a
3-D pre-stack depth migration. They showed that 2-D analysis produced artifacts while
the 3-D algorithms provided a much more accurate picture.
In this work, we address the survey design issues of 3-D VSP via a 3-D VSP CDP
binning procedure. We also propose procedures for handling and processing of model
and field 3-D VSP data. In this numerical modeling paper, we will not distinguish
between VSP and reverse VSP procedures.
Data set B is from a previous study (Sun and Stewart, 1994). The model consists of
three layers (Figure 1b). The thicknesses of the three layers from top to bottom are 800
m, 400 m and 800 m respectively. A dome centered at (1500m, 1500m) is located on
the top of the second layer with a radius of 630 m and height of 210 m, which
possesses dip angles varied from 0 to 36.9. The dataset was generated with
reflections only from bottoms of the first layer (dome structure) and second layer.
The source and receiver geometry configuration of both datasets are the same, and
are shown in Figure 2. The examples of raw records of both shots gathers and receiver
gathers are shown in Figure 3.
w ell w ell
E
surface
surface R R
S layer 1
layer 1 V p = 4000 m/s S V p = 4000 m/s
z=800 z=800
layer 2
V p = 5200 m/s layer 2 V p = 5200 m/s
z=1100-1400
z=1200
layer 3
V p = 6000 m/s layer 3 V p = 6000 m/s
z=2000 z=2000
(a) (b)
X_ COORD
FIG. 2. Surface receiver layout geometry (61 receiver lines with a 50 m line spacing and 61
receivers on each line with 50 m receiver spacing). The well location is shown as a dot in the
centre of the receivers area.
In the 2-D case, the lateral offset of the reflection point Xr (in Figure 4a) is given
by:
Xr = X
2
1 + V2R TR , (1)
VRBTRB
where VR is the RMS velocity of upgoing raypath or downgoing raypath , VRB is the
RMS velocity for downgoing raypath or upgoing raypath, TR is the zero-offset
traveltime (one way) from surface to reflection point; and TRB is the zero-offset
traveltime (one way) from reflection point to borehole a source (for RVSP) or a receiver
(for VSP). X is the source-receiver offset.
The concept used to calculate the reflection point of 3-D VSP is diagrammed in
Figure 4. The reflection point denoted by C in Figure 4(b) is in the xy plane instead of
in x direction only .The reflection point is located in the propagation plane of the P
wave. We can define the azimuth of the reflection point by calculating the source-
receiver line azimuth angle with the X or Y axis; then is written as
Y 1
= tan (2)
X ,
where X and Y are the horizontal offsets from the well top in the y and x direction.
X + Y
2 2
Roffset = 2 (3)
VR TR ,
1+ 2
VRBTRB
where X and Y are the horizontal offset from the surface location to the well top in
the y and x direction.
The horizontal distance covered by the reflected wave in the x direction is given by
Xr =Roffsetsin (4)
The horizontal distance covered by the reflected wave in the y direction is given by
Yr =Roffsetcos
(5)
Equation (3) provide the binning method. Once we calculate the coverage of the
reflection point by using equation (3), we can split it into the x and y axes by using
equations (4) and (5).
Since the shots are in a two dimension plane, the 3-D VSP has the character of
surface 3-D seismic data. Therefore, it is necessary to define the bin grid in two
dimensions as in surface 3-D data binning. This is shown in Figure 5a. On the other
hand, the receivers in the well borehole, the 3-D VSP also has the character of VSP
seismic in which is the reflection points for each recorded trace are variant in depth
domain as shown in figure 5b. If we define the same bin origin and grids in 2-D plane
for all depth layers of interests, for same trace at different depth points, the reflection
points will be located in different bin cell.
By combining the two characters of 3-D VSP, the bin grids are defined in a two
dimensional plane, and the reflection points are calculated in 2-D plane at each depth
point by using equations (3) to (5) and the reflection points of the 3-D VSP are mapped
in 3-D position.
Well
X R
Layer 1
Layer M S i
Layer i
Layer N
Xr (a)
Y
R
Well i
C
(b)
FIG. 4. VSP ray geometry for a flat horizontally layered medium , (a): 2-D case; (b): 3-D case.
Y
Receiver
Crossline
Bin cell
Borehole
y
X
x
(a) Inline
x Receiver
0 X
Bin origin
Shot
Depth
Layer i
Reflection point
(b)
FIG. 5. Schematic diagram showing binning algorithm. (a) surface diagram; (b): depth
domain diagram
Figures 6 and 7 are used to illustrate the binfold distribution as depth (time)
increases. These two results are the binning algorithm performed on the same dataset
where the inline and crossline receiver coordinates are from 200 to 2800 meters, the
inline and crossline receiver intervals are 50 meters. There are 30 shots located at the
surface 1500 m in both directions, and located from surface down to the well to
borehole 1200 m with depth interval 30 meters. All the shots are above (one is on) the
second layer in the model A, which means the results in Figure 6 and 7 based on the
same surface and depth geometry.
We use same bin interval, azimuth and bin origin on these layers. the bin origin and
azimuth are the same as surface layout coordinate origin and azimuth. the bin interval
in both directions is 15 m.
Figure 6 shows the areal bin fold distribution (Figure 6a) around 600 ms (the
reflection of the second layer in model A). The bin fold statistics are shown in the
histogram (Figure 6b). The overall the fold distribution is not quite uniform. Two
"black" strips across the covered area East-West and North-South through the Well
with low fold coverage (from 3 fold to 40 fold) along which higher fold coverage (60
to 70 fold) are surrounded (Figure 6a). Two "bright" strips across the covered area
through the Well 45 from any of the "black" strips with high fold coverage (80 to 100
fold).The dominant coverage is about 50 to 60 fold but they are distributed in the 45
azimuth of receiver line direction. There also have quite big portion of bin cell with low
fold (3 to 10 ) coverage distributed along the first and last 10 inline and crossline. As
the depth (time) increases, the bin fold distribution gets more uniform as illustrated in
Figure 7. As in figure 6, there are two "black" strips across the well top in receiver
inline and crossline direction. The major fold distribution becomes much more uniform
and this also can be viewed in figure 7(b),the histogram shows better distribution, and
much higher fold. The dominant fold coverage of the third layer is around 40-75 fold,
and there is almost no low fold less 35 between 10-80 inline and crossline of CRPs
(reference numbers) as displayed in figure 15 except the two "black" strips cross the
well top.
Comparing Figures 6 and 7, we find out that as the depth increase the bin fold will
distributed more uniformly and is higher than in shallower layers.
Figure 8 shows the bin fold distribution of the same dataset of layer 3 by using same
parameters except now shifting the bin origin 7.5 m towards east and north directions.
We can see the two "black" strips have disappeared. Figure 9 shows the bin fold
distribution of layer 3, the bin size of 30 m in both direction. Comparing Figure 9 and
7, we can see by increasing the bin size, the binfold will distribute more uniformly with
high fold.
Figures 10 illustrate the fold distribution when few shots are used. In this Figure,
we take out every second shot as shown in Figure 7, giving only 15 shot, They located
from the surface down to the well to 1160 m with an 80 m interval. Comparing Figure
10 with Figure 7, we find that the binfold distribution is not much different. The
dominate binfold is 22 to 27, which is half amount of its shown in Figure 7. This
indicates that the number of shots or receivers downhole does not strongly influence the
fold distributions.
Figure 11 illustrates the fold distribution when the surface geometry changes. In this
test, we keep the shot depth and number the same as in Figure 7, but we take out every
second receiver in the surface ,giving a receiver interval 100 meters in both inline and
crossline. The result in Figure 11 shows the binfold is low and there are lot of zero bins
inside the offset 700 meters. This tells us the surface station geometry heavily
influences both the bin size and the bin fold.
Through these tests, we find that the minimum bin size is 1/3 inline interval * 1/3
crossline interval
t = TR + TRB + X2 (6)
2 2 ,
2 V T + VRBTRB
R R
Extending equation (6) to the 3-D domain, the traveltime of the P-P reflected wave
from the source to the receivers is written as
2
X + Y
2 2
t = TR + TRB + , (7)
2 2
2VT +V T
R R RB RB
where X and Y are the horizontal offset from the surface location to the well top in y
and x direction.
Equation (7) provides the NMO correction method for 3-D VSP wavefield. That is
to put the amplitude of seismic wavefield at t time to a two-way normal incidence P-P
time (TR) in two dimensional domain.
In this paper, the NMO algorithm is performed on two numerical 3-D VSP model
datasets. The procedure is shown in Figure 12. The input data is the preprocessed
upgoing wave. RMS velocities and the raypath travel time are calculated by using
known interval velocity and depth information.
Figure 14 shows some of the upgoing waves of model A and B. Three receiver
gathers from each model are located at north-southing 1500 m (cross well top) and east-
westing 1500,2000, 2700 meters, the shot depth are from 1200 meters down the
borehole up to the surface. The preprocessed upgoing wavefield is the input into the
NMO processing flow.
Figure 15 shows some of the results of NMO applied crossline section of model A.
Data are sorted in common shot gathers at different surface inline locations and variant
shot depths. Figure 15a and 15b are the same shot gather at shot depth at surface, but
the inline location of the receivers are located at 1500 meters ( at the well top), and 2200
meters (700 meters away from the well top) differently. Comparing these two, we can
see as the horizontal offset of receiver increase, the accuracy of NMO correction on the
event at the far offset of the shallow layer decrease. Figure 15c is a shot gather located
at the some location of figure 15a but the shot depth is 120 meters down the well
borehole. Comparing this two, we found out as the shot depth becomes deeper, the
error grows bigger. Figure 9d is the inline receiver located at 50 meters away form the
well top, but the shot depth at 320 meters down the borehole. This figure have proved
the effect at the far offset at the shallow layer.
Figure 16 shows the result of NMO applied on dome model B. The geometry is the
same as described (in Figure 15).We can see from this Figure, that the NMO algorithm
performed well, except at shallow layer at far offset, the deeper shots depth (320 m in
the figure 10d).
Over all, particularly the NMO correction performed well on the deeper layers of
both dipping layered model and dome structured model.
Figure 18a and 18b are inline common reflection point gather (CRP) stack section
(east-west direction) for the dipping layered model A. Figure 18a is the CRP bin
located at crossline near offset 1515 meters, which is 15 meters away from the well top
horizontally. Figure 18b is the CRP bin located at crossline far offset 1930 meters ,
which is 430 meters away from the well top. These two figures show that the binning
algorithm performed well. On the first layer (400 ms in time) there is not much data,
that is because the error of NMO correction.
Figure 19 shows some of the crossline CRP stack sections. Figure 19a shows the
inline CRP bin at 1530 meters, which is 30 meters away from well. Figure 19b shows
the inline CRP at 2025 meters, which is 525 meters away from well top. These figures
indicated the binning and stacking algorithm works reasonably well.
Figure 20a and 20b show some CRP stacks from model B. Figure 20a is located at
inline CRP bin coordinate of 1515 m while 20b is at 1230 m. In areas of high dip there
are some mapping errors. This is due to mapping velocity errors and the flat layer
assumptions. We expect that the depth migration will migrate the structure to it true
location and this will be part of our future work.
CONCLUSIONS
In this study, we have developed a 3-D VSP NMO and binning algorithm. The
raypath NMO traveltime performs well when the horizontal offsets were less than 1200
meters for the dipping layered model A, and 1000 meters for structure layered model B
at shallow layer. When the offset is too large or the velocity at the shallow layer is too
slow, a large error will occur in the mapping.
3-D VSP binning has both characteristics of VSP and surface 3-D seismic.
Surface geometry is used to determine the bin size and bin azimuth. The surface
intervals are the major influence on the bin fold. The number of downhole positions is
the major parameter for the bin fold. Changing the number of the downhole locations or
the tool interval has no much influence on the distribution of the bin fold.
FUTURE WORK
There are numerous aspects of 3-D VSP survey design to investigate. For example,
if we have a limited receiver array, where should it be positioned to give the most
broad, but uniform fold coverage? In processing, there are many algorithms to adapt to
the 3-D VSP case. We need to be able to do velocity analysis (from direct arrivals and
reflections ) as well as ray trace mapping and prestack migration
ACKNOWLEDGMENTS
The authors would like to acknowledge Darren S. Foltinek for his assistance with
coding the binning program in Promax.
REFERENCES
Aleotti, L., Gallori, A., Miranda, F., Craglietto, A., Persoglia, S., Poletto, F., Impact of drill-bit
seismic method on exploitative wells: Presented at the 56th Ann. Mtg., Europ. Assn. Explor.
Geophys., Vienna.
Chen, H. and McMechan, G.A., 1992, 3-D pre-stack depth migration for salt and subsalt structures
using reverse-VSP data: J. Seis. Expl., 1, 281-291.
Eaton, W. S. D., Stewart, R. R., and Harrison, M. P., 1991, The Fresnel zone for P-SV waves:
Geophysics 56, 360-364.
Haldorsen, J. B. U., Miller, D. E., Walsh, J. J., and Zoch, H.-J., 1992, Multichannel approach to
signature estimation and deconvolution for drill bit imaging: Presented at the 62nd Ann. Intl.
Mtg., Soc. Explor. Geophys., New Orleans.
Hardage, B.A., 1992, Reverse VSP and crosswell seismology: Geophysical Press.
Jackson, P.J., Onions, K.R., and Westerman, A.R., 1989, Use of inverted VSP to enhance the
exploration value of boreholes: First Break, 7, 233-246.
Jardine, D. 1974, Cretaceous oil sands of Western Canada: CSPG Memoir 3, 50-67.
Jones, M., 1991, On the analysis of a reverse VSP data set using a core-gun source: Presented at the
Sixty-First Ann. Intl. Mtg. Soc. Expl. Geophys., Houston.
Labont, S., 1990, Modal separation, mapping, and inverting three-component VSP data: M.Sc.
thesis, University of Calgary.
Kragh, J.E., Goulty, N.R., and Findlay, M.J., 1991, Hole-to-surface seismic reflection surveys for
shallow coal exploration: First Break, 7, 335-344.
Minken, F.D., 1974, The Cold Lake Oil Sands: Geology and a reserve estimate: CSPG Memoir 3,
84-89.
Naville, C., Layotte, P.C., Serbutoviez, S., and Verdier, F., 1991, Uphole surveys digitally recorded
and processed as multi-offset RVSPs: Presented at Ann. Mtg. Europ. Assoc. Expl.
Geophys., Paris.
Parra, J. O. and Bangs, J. H., 1992, High-resolution reverse VSP and interwell seismic experiments at
the Buckhorn test site in Illinois: Presented at the 62nd Ann. Intl. Mtg., Soc. Explor.
Geophys., New Orleans.
Stewart, R. R., 1991, Rapid map and inversion of P-S waves: Geophysics, 56, 859-862.
Sun, Z. and Stewart, R. R., 1994, 3-D reverse VSP: CREWES Report, 1994.
Toksz, M. N. and Stewart, R. R., 1984, Vertical seismic profiling, Part B: Advanced concepts:
Geophysical Press, 1984.
10
20
30
Inline bin number
40
50
60
70
80
90 10 20 30 40 50 60 70 80
( b)
Count
11 5
5 0000
1100 0 0
5 000
50
10
20
Inline bin number
30
40
50
60
70
80
(b) (b)
2000
Count
1000
0
0 37.5 75 112.5 150
FIG. 7. Bin fold distribution display of the third layer for the full dataset of model A. (a): Areal
fold distribution, (b): histogram of fold distribution.
10
20
30
Inline bin number
40
50
60
70
80
0 10 20 30 40 50 60 70 80
(b)
Count
FIG. 8. Bin fold distribution display of third layer of full dataset of model A, the receiver interval
is 50m, shot interval: 40m, bin size 15 *15 with a bin shift of 7.5 m towards north and west
direction. (a): Areal fold distribution, (b) histogram of fold distribution
10
15
Inline bin number
20
25
30
35
40
5 10 15 20 25 30 35 40
(b)
FIG. 9. Bin fold distribution display of third layer of full dataset of model A. (a) : Areal fold
distribution, (b): histogram of fold distribution. Receiver interval 50 m, shot interval: 40m, bin
size 30*30 m.
10
20
Inline bin numble
30
40
50
60
70
10 20 30 40 50 60 70
(b)
1500
1000
500
0
0 20 40 60 80
FIG 10. Bin fold distribution display of the third layer of model A. All receivers, with a receiver
interval of 50m, 15 shots with an interval of 80m, bin size: 15*15m. (a): Areal fold distribution,
(b): histogram
10
20
30
Inline number
40
50
60
70
80
90
10 20 30 40 50 60 70 90
(b)
2000
1000
0
0 37.5 75 112.5 150
FIG. 11. Bin fold distribution using one quarter of the surface receivers. Receiver interval of
100m , 30 shots with a 40 m interval. The bin size is15*15m. (a): Areal fold distribution; (b):
histogram of fold distribution.
FIG. 13. 3-D VSP P-P wave binning and stacking flowchart
34-20
0 0
Zhang ,Stewart and Sun
500 500
Time(ms)
800
300
400
500
Time(ms)
600
700
900
(a) (b) (c) (d)
FIG. 15. The NMO applied section on dataset A in common shot gather. Shot depth at : (a): surface;
(b):surface; (c):120 m; (d): 320 m.
34-21
3-D VSP Processing
34-22
1500 2200 1500 1550
200 1200 2200 550 1550 2550 400 1400 2400 250 1250 2250
200
200
Zhang ,Stewart and Sun
300 300
400 400
Time(ms)
500
500
600
600
700 700
400
Time(ms)
600
800
1000
(a) (b)
FIG. 17. The binning and stack section of model A. (a): the Inline section at cdp_x 1530 m,
(b):the Inline section at cdp_y 1935 m.
400
Time(ms)
600
800
1000
FIG. 18. The binning and stack section of model A. (a): the Cross-line section at cdp_x 1530
m, (b): the cross-line section at cdp_x 2050 m.
300
400
Time(ms)
500
600
700
(a)
Rec_y 910 1110 1310 1510 1710 1910 2110
200
300
Time(ms)
400
500
600
700
(b)
FIG.19. The binning and stack section of model B. (a): At near offset cdp_x 1515 m; (b): At far
offset cdp_x 1230 m.