GEOPHYSICS, VOL. 68, NO. 5 (SEPTEMBER-OCTOBER 2003); P. 16951707, 19 FIGS.
10.1190/1.1620643
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Reduced-time migration of transmitted PS waves
David Sheley and Gerard T. Schuster
ABSTRACT
We develop the novel theory of transmitted PS migra-
tion and show that PS transmitted arrivals in a Gulf of
Mexico vertical seismic prole (VSP) data set can be mi-
grated to accurately image a salt sheet even though the
receiver array is below the transmitting boundary. We
also show that migrating transmitted arrivals is effective
in illuminating the base of an orebody invisible to PP
reections. In general, interfaces that bisect wavepath
propagation (i.e., the source and receiver are on oppo-
site sides of the interface and therefore invisible to PP
reections) can be imaged by migration of PS transmit-
ted waves. These results suggest that migration of PS
transmitted waves opens new opportunities in imaging
nearly vertical impedance boundaries that are typically
invisible to conventional reection imaging of crosswell
and VSP data.
We also present a new interferometric method,
denoted as reduced-time migration, which uses the
arrival-time difference between the direct P-wave and
subsequent events to increase migration accuracy.
Reduced-time migration removes static time shifts in
INTRODUCTION
Seismic migration is a powerful imaging method (Gardner,
1985) used to reconstruct the reectivity distribution from seis-
mic reection data. Originally developed for common-depth-
point (CDP) data, it was extended in the 1980s and 1990s to
image the reectivity distribution from both vertical seismic
prole (VSP) (Amundsen et al., 1993; Payne, 1994) and cross-
well data (Hu et al., 1988; Qin, 1994). Although migration of
PP reections is effective in imaging horizontal or obliquely-
dipping boundaries, it is notoriously ineffective in delineat-
ing near-vertical interfaces such as salt anks or orebody
boundaries.
The inability to image steep anks is a serious economic lia-
bility because oil or minerals are often adjacent to such bound-
Manuscript received by the Editor September 4, 2001; revised manuscript received July 5, 2002.
University of Utah, Department of Geology and Geophysics, WBB Room 719, 1460 East 135 South, Salt Lake City, Utah 84112-0111. E-mail:
dwshele@ppco.com; schuster@mines.utah.edu.
c 2003 Society of Exploration Geophysicists. All rights reserved.
1695
the data, decreases the focusing error due to an incor-
rect migration velocity model, and relocates reection
or PS transmission events to be closer to their true
positions. Although limited to crosswell and VSP ge-
ometries, synthetic- and eld-data examples show that
reduced-time migration is noticeably more accurate than
conventional migration in the presence of static shifts
and/or migration velocity errors. The main assumption
of reduced-time migration is that the direct wave samples
errors which are representative of errors in the migration
aperture. Transmission wavepaths, in general, are sub-
parallel to the direct wave and therefore the two modes
encounter similar errors and, hence, reduced-time mi-
gration is effective in improving the focusing of migra-
tion energy. For the PP reection case, the direct wave
and the reected waves often traverse different parts of
the earth, therefore, reduced-time migration will remove
static shifts but it is not expected to mitigate velocity er-
rors if the errors are spatially variant. However, if there
is a general and consistent bias in the velocity model,
reduced-time migration is expected to deliver improved
results over conventional Kirchhoff migration.
aries. To solve the steep boundary problem we propose migrat-
ing the forward-scattered PS-waves to locate the transmitting
boundary. Previous work on forward-scattering by Balch et al.
(1991) showed that reverse-time migration of converted waves
could image point diffractors in a physical model. This led to
the further development of PS-reection migration of cross-
well data, but the imaging of PS transmitted waves remained
unexplored.
The earthquake community also uses PS arrivals to im-
age certain layer boundaries. Using a layered model assump-
tion, earthquake seismologists use P and S transmitted waves
from earthquakes to image the Moho, denoting the method
as the receiver-function technique (Langston, 1977). Later,
Chavez-P
erez
(1997) and Chavez-P
erez
and Louie (1996, 1998)
used forward-scattered P-waves from recorded earthquakes to
 1696 Sheley and Schuster
infer the presence of deep fault-zone diffractors. Chavez-P
erez
(1997) proposed using either forward-scattered S-waves or PS
events to image diffractors, but results with earthquake data
were arguably ambiguous.
Although these previous researchers used PS transmission
data, they did not exploit its most useful feature: namely, imag-
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
ing vertical interfaces with VSP or crosswell data. We expand
on this previous work and dene a Kirchhoff migration oper-
ator for transmitted converted waves and use it to successfully
image transmitting boundaries in both synthetic and eld data.
This overcomes, to a large degree, the inability of seismic meth-
ods to image vertical boundaries.
Other major problems in migration include an incorrect mi-
gration velocity model and static shifts in the data. Since such
high frequencies are used, static shift problems are particularly
acute in the mining industry, where errors in source initiation
time or indeterminate well locations lead to defocusing of the
migration image. To mitigate these problems, we introduce a
new interferometric technique that decreases the sensitivity of
migration to velocity model errors and removes static errors.
We denote this method as reduced-time migration. Reduced-
time migration is validated with both synthetic and eld data,
and we show that it can be applied to both transmission and re-
ection arrivals to improve image quality. By combining these
two tools, transmission migration and reduced-time migration,
several of the major problems that have long plagued crosswell
and VSP migration are now mitigated.
The rst two sections of this paper describe respectively the
theory of PS transmission migration and the theory of reduced-
time migration; following this, we present some numerical and
eld data results. The last section offers some conclusions.
TRANSMISSION PS MIGRATION THEORY
Converted-wave migration migrates either reected or
transmitted mode-converted events to their place of conver-
sion at the reecting or transmitting boundary. Compared to PP
reection migration, the advantage of migrating transmitted
converted waves in crosswell and VSP surveys is that bound-
aries roughly parallel to the wells can be imaged, as illustrated
in Figure 1. To demonstrate this capability, Figure 2 shows sev-
FIG. 1. Division of energy from an incident P-wave. The trans-
mitted arrivals are extant in the receiver well, no reections
can be recorded.
eral migration isochrons, the curves along which PP reection
(Figure 2a) and converted PS (Figure 2b) amplitudes will be
smeared by Kirchhoff migration. Each technique is suitable
for imaging boundaries tangent to the migration isochrons.
For a single trace and a homogeneous medium, PP reection
migration smears reection amplitudes along isochrons that
form concentric ellipses, whereas, PS migration smears trans-
mission amplitudes along isochrons that form oblate ellipsoids.
The region near the tip of the oblate ellipsoids (dashed lines
in Figure 2b) is where the P-waves convert to transmitted S-
waves, which is the event we exploit in imaging vertical bound-
aries. Note, there are no vertical tangents between the wells for
the PP reection isochrons, meaning that such boundaries are
invisible to PP reection imaging. Contrast this with the many
vertical or near-vertical tangents for the transmitted PS-wave
isochrons.
Similar to PP-reection migration, the migration algorithm
for transmitted PS arrivals, in a single trace of crosswell or VSP
data, consists of three steps: (1) calculate the traveltime srP for
P-waves to propagate from the source to an image point using
the P-wave velocity, (2) calculate the traveltime rgS for S-waves
to propagate from an image point to the receiver using the
S-wave velocity, and (3) smear the PS transmission amplitude
from the trace at time equal to srP + rgS to the image point
denoted by r:

 
m(r) = h(s, r, g)s z g , srP + rg
S
dz g , (1)
FIG. 2. Comparison of (a) PP reection migration isochrons
and (b) PS converted-wave isochrons in seconds. The PP
isochrons form ellipses with the source and receiver as foci,
whereas the PS converted-wave isochrons form oblate ellip-
soid (egg-shaped) isochrons. The tangents to the dashed por-
tions of the isochrons denote the interfaces that can give rise to
transmitted PS arrivals; solid lines indicate those for reected
waves. A transmission ray is shown in green and reections in
red.
 Migration of transmission PS waves 1697

where m(r) is the migration image at r, s(z g , t) is the seismic els. Figure 2b shows the isochrons for srP + rgS , the imaging
trace recorded at the depth denoted by z g , and condition for both PS reection (solid lines) and transmis-
 sion PS-waves (dashed lines). If rgS rgP , then equation (1)
1 if sr and rg form a transmitted ray, is appropriate for PP reection migration, with the associated
h(s, r, g) =
0 otherwise, isochrons shown in the Figure 2a.
The importance of Figure 2 is that the tangents to the
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

is the threshold lter that distinguishes transmitted rays transmission PS-wave isochrons are vertical or nearly vertical,
from reected rays. Migrating the data in this manner re- whereas those for PS reected waves or PP reections have
quires knowledge of both the P- and S-wave velocity mod- less dip between the wells. This means that PS transmission

FIG. 3. (a) Homogeneous velocity model with source-receiver spacing of 1000 m and a depth of 200 m. The depth
of reectors R1 and R2 is 759 m and 1066 m, respectively. (b) Isochrons are governed by srP + rgP . The direct wave
(D) migrates to the line connecting the source and receiver. (c) Migration isochrons for (a) but the traces have
been shifted by 0.2 s prior to migration. (d) Same as for (c) but reduced-time migration was used. (e) Migration
isochrons for a 110% velocity model. (f) Same as for (e) but reduced-time migration was used.
 1698 Sheley and Schuster
migration is appropriate for imaging steep dips compared to
the limitation of imaging only shallower dips by PP and PS
reection migration. A resolution analysis of PS transmission
migration is given in Appendix A, which shows that resolution
increases as the P- to S-velocity contrast increases.
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
REDUCED-TIME MIGRATION THEORY
Reduced-time migration is dened as temporally shifting the
traces by the traveltime of the observed direct arrival sg
obs
, and
then migrating the reduced data, i.e.,

 
m(r) = s z g , sr + rg + sg
obs
sg
calc
dz g ,
where sgcalc
is the direct traveltime computed from the velocity
model. This sg calc
term is used to adjust the migration operator
to account for reduced-time data. If the exact velocity model is
known, then sg obs
= sg
calc
, and this equation does not differ from
the conventional Kirchhoff migration equation. However, if
there is an unknown source static time shift tsg in the traces,
then sg
obs
sgcalc
= tsg , which is the negative value of the un-
known static shift. Thus, equation (2) eliminates this static error
by reshifting the traces by tsg .
The focusing error due to a static shift in the trace is illus-
trated in Figure 3. Figure 3a shows the model where a single
trace is assumed to contain three events: a direct P-wave de-
noted as D, a reection R1 from layer one, and a reection R2
from layer two. This trace is migrated to give the section shown
in Figure 3b, where the direct wave energy has been smeared
along a line connecting the source with the receiver, and reec-
tion energy is smeared along the correct ellipses. If the trace
FIG. 4. Vertical boundary model used to generate synthetic
crosswell and RVSP data. Density is constant for the entire
100-m2 model, and a 1500-Hz z-component source was used.
Sources for both crosswell and RVSP simulations are on the
left at 1-m intervals (101 total). Receivers (101) at 1-m intervals
are distributed along the right for the crosswell simulation, and
across the top for the RVSP simulation.
is now given an erroneous static shift tsg , the direct wave D
is now incorrectly smeared along the jagged ellipse shown in
Figure 3c, and the reections are also smeared along incorrect
ellipses. However, reduced-time migration eliminates this er-
ror as shown in Figure 3d because the term sg obs
sg
calc
= tsg
compensates for the unknown static shift.
Reduced-time migration also mitigates errors due to erro-
neous migration velocities. In the case of a migration velocity
error, the erroneous time shift to the data is in the traveltime
calculation of the imaging condition, where
sr + rg = srcorr ect + rg
corr ect
+ tsrg . (3)
(2)
Here, the correct superscript refers to the calculated travel-
times in the correct model, and tsrg is the error in comput-
ing sr + rg with an incorrect velocity model. Note, if r is a
point on the direct wave ray, then sg obs
sg
calc
= tsrg in equa-
tion (2) and the direct wave is migrated correctly along the di-
rect ray. However, if r is elsewhere, then sgobs
sgcalc
 = tsrg , so
the reectors are still migrated incorrectly. But, the associated
static shift error is reduced because |sg
obs
sgcalc
| < |tsrg |. This
is illustrated in Figure 3, where the correct migration image is
given in Figure 3b, whereas migration with an incorrect velocity
model results in Figure 3e. Note, the direct wave is now incor-
rectly migrated to an ellipse, and the reections are migrated
deeper than their reecting boundaries. With the incorrect ve-
locity model, reduced-time migration (Figure 3f) migrates the
direct wave D to its proper position and migrates the reection
FIG. 5. Synthetic RVSP seismogram for a source at 100-m
depth. The source frequency is 1500 Hz, receiver spacing is
1 m, and the total record length is 0.06 s.
 Migration of transmission PS waves 1699

events R1 and R2 closer to their reecting boundaries than with Vertical boundary model
conventional migration. This is because the compensation fac-
tor sg
obs
sg
calc
in equation (2) reduces the value of the velocity A vertical boundary model is used as a feasibility test and to
static shift factor tsrg in equation (3). gain insight into both PS transmission migration and reduced-
time migration. Both crosswell and reverse vertical seismic pro-
NUMERICAL TESTS le (RVSP) data are generated and migrated.
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

In this section, synthetic data are generated for a vertical

boundary model and a salt diapir model. The transmitted PS Forward modeling.The vertical boundary model in
arrivals are then migrated to image the converting boundary, Figure 4 is used to generate both crosswell and RVSP data.
demonstrating the ability of PS-transmission migration to im- P-wave velocities are 5000 and 5500 m/s, the P- to S-wave ve-
age such features. locity ratio is 1.5, and a constant density is used. Parallel source

FIG. 6. Migration images for the crosswell experiment. The sources are on the left and the receivers on the right.
The PS transmitted waves were isolated by muting. (a) Migration with the actual velocity model. (b) Shows the
conventional migration result with a 90% velocity model. (c) In the presence of a 10% velocity error, reduced-time
migration places the transmitting boundary much closer to its true position than does conventional migration.

FIG. 7. Migration images for the RVSP experiment. The sources were on the left and the receivers were along
the top, both emplaced at 1-m intervals. Note that only a small portion of the vertical boundary is imaged due to
the restricted RVSP geometry and refraction effects.

FIG. 8. SP migration for the RVSP experiment. Note that much more of the transmitting boundary is imaged
than with PS transmitted waves (Figure 7).
 1700 Sheley and Schuster

and receiver wells bracket the model for the crosswell sim-
ulation. The RVSP simulation uses the same sources as for
the crosswell simulation but the receivers are along the top
of the model. Synthetic seismograms were computed using a
2D elastic wave modeling code, placing sources and receivers
at 1-m intervals. The model grid size is 20 cm, the source fre-
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
quency is centered at 1500 Hz, and the time interval is 10 s.
The line source is orthogonal to the model space and, there-
fore, geometric spreading occurs in only two dimensions. Both
vertical- and horizontal-component data were generated and
recorded. A common-shot gather for a source at 100-m depth
and receivers on the surface (RVSP) is shown in Figure 5.

FIG. 9. Migration images for reected PP events in the RVSP experiment.

FIG. 10. Kidd Creek P-wave velocity model assuming a 3.5-ms time delay. The inversion data residual is 0.086 ms.
The source-well (right) and receiver-well (left) locations are indicated by the solid black lines. The dashed lines
represent the boundaries of the orebody as inferred from well information.
 Migration of transmission PS waves
Seismograms from both crosswell and RVSP data sets are
then migrated using a Kirchhoff migration algorithm (French,
1974; Gardner et al., 1974) in combination with reduced-time
migration.
Migration results.Prior to migration PS transmission
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
events were isolated by muting all other arrivals. Using no re-
striction on incidence angles, Kirchhoff PS migration was ap-
plied to the PS events for both the true velocity (Figure 6a)
and for a 90% velocity model (Figure 6b), i.e., 90% of
V p1 , V p2 , V s1 , and V s2 . The PS events are migrated to their
actual position for the true velocity model and to an incor-
rect position for the 90% velocity model. Reduced-time migra-
tion was also applied to the transmission PS arrivals. Results
for reduced-time PS migration are shown in Figure 6c, where
the location of the transmitting boundary for the 90% velocity
model is much closer to the true boundary position than with
conventional migration.
The results for synthetic RVSP data are similar to those
for crosswell data. Figure 7 shows the results for conventional
and reduced-time migration of transmitted PS-waves. Figure 8
shows the results for transmitted SP-waves. In addition to the
transmitted events, the reected PP-waves were migrated by
Kirchhoff (Figures 9a and 9b) and reduced-time migration
(Figure 9c).
Note, much more of the boundary is imaged by migrating
the SP-waves than by migrating the PS- or PP-waves. For this
FIG. 11. Common receiver gather for a receiver at a depth of
20 m. The gather has been bandpass ltered between 1000 and
6000 Hz.
1701
model, the velocity contrast for SP-waves is greater than that
for PS- or PP-waves; hence, the SP-waves are refracted more
and give a wider target illumination.
FIELD-DATA RESULTS
Applying converted-wave and reduced-time migration to
crosswell and a VSP eld data conrm the vertical boundary
imaging of PS transmission migration. Converted-wave migra-
tion illuminates the base of an orebody and the top of a salt
sheet, both invisible to conventional migration of PP reec-
tions. Also, an unknown static shift in the crosswell example is
eradicated by reduced-time migration.
Kidd Creek crosswell imaging
The Kidd Creek mine, near Timmins, Ontario, Canada, is
a volcanogenic sulde deposit consisting primarily of rhyolite,
basalt, diorite, ultramacs, argillite, and massive suldes. Previ-
ous work proved the feasibility of imaging massive sulde de-
posits with seismic reections (Meng and McGaughey, 1996).
Here, we migrate both PP reection and SP transmission events
to image the boundaries of the sulphide body embedded in the
host rock. We also show the benets of applying reduced-time
migration to these data.
FIG. 12. The gather in Figure 11 has been attened to the direct
P-wave traveltime and median ltered, and the direct S-wave
and later events have been muted.
 1702 Sheley and Schuster

Data description.The Kidd Creek crosswell data were col-
lected by Noranda Mining and Exploration using boreholes
drilled from inside the copper-zinc sulde mine, and consists
of 140 shots with 48 receivers each. The boreholes intersect an
orebody, as evidenced by cores from each hole. These seismic
data are ideally suited for the reduced-time migration method
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Migration results.Prior to migration, the data were pro-
cessed by resorting into the common receiver gather (CRG)
domain, picking the rst-arrival traveltimes, attening the
direct-P arrivals, and then picking the residual times to further
atten the direct P-wave. After attening, an 11-point median
lter was applied to accentuate the reection and transmission

since there exists an unknown time delay and/or well location
error in the data. To mitigate the problem, Hu (1999) proposed
that a constant time shift be applied to the traces. He estimated
this delay to be 3.5 ms by assuming a constant background ve-
locity and minimizing the data residual. After applying the data
time shift, the rst arrival traveltimes were inverted to gener-
ate the velocity model shown in Figure 10. Since the velocity
tomogram does not correlate well with the log-based model,
the tomogram is assumed to be in error. Nevertheless, it is
used here as the migration velocity model to prove the util-
ity of reduced-time migration. Since the S-wave rst arrivals
are not distinct throughout the data set, they were not picked
and inverted to generate an S-wave velocity model. Instead,
we assume a constant P-to-S velocity ratio of 1.7. This value is
consistent with empirical data from an adjacent mine.
events; all but these events were then muted and unattened
back to their original positions. A sample gather after bandpass
ltering is shown in Figure 11, and Figure 12 shows the result
after both the bandpass and median ltering.
By applying conventional converted-wave migration to the
muted and time-shifted CRGs, the transmitted SP-waves,
which appear as PS transmissions in the CRGs, are migrated
near the orebody boundary estimated by Noranda (Figure 13a).
A restriction on incidence angles was used to limit the angle be-
tween the image point and the line connecting the source and
receiver to be less than 20 . That is, the P- to S-conversion takes
place mostly along the tips of the oblate ellipsoids in Figure 2.
This restriction helped to reduce migration artifacts.
The reduced-time migration algorithm was then applied to
the seismograms without applying the 3.5 ms time shift. The

FIG. 13. Result from migrating the SP transmission events from the bottom of the orebody. Eight CRGs were
migrated and stacked to produce this image; each CRG contained 140 traces. (a) Conventional SP transmission
migration image. (b) The same data for (a) were migrated with reduced-time migration. (c) Result from migrating
the PP reection events from the top of the orebody with a convention migration algorithm. Seven CRGs were
migrated and stacked to produce this image. (d) The same data for (c) were migrated with the reduced-time
migration equation for PP reections.
 Migration of transmission PS waves 1703

result, given in Figure 13b, migrates the events much closer
to the inferred boundary than with conventional migration.
Moreover, the intersection of the migrated boundary with the
borehole and the intersection of the SP transmitted wave with
the direct P-wave correlate well.
Hu (1999) used conventional PP reection migration to im-
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
Gulf of Mexico VSP imaging
We now describe the results of applying PS transmission mi-
gration to offset VSP data acquired in the Gulf of Mexico. The
objective is to use PS transmission events to image the top and
base of a tabular salt sheet.

age the top boundary of the ore body. We repeat the pro-
cess here and compare it to PP reection migration using the
reduced-time imaging condition. We found a strong reection
event from the top boundary in seven CRGs. These gathers
were processed, as described in the previous section, by me-
dian ltering and muting. After applying the 3.5-ms data shift
to the traces, the gathers were migrated using PP reection
migration (Figure 13c). Without the time shift, the reection
energy is either not migrated or is smeared far from its location
on the ore boundary.
Reduced-time migration was then applied to the unshifted
data. The result, given in Figure 13d, migrates the events much
closer to the inferred boundary than with conventional mi-
gration. Moreover the intersection of the boundary with the
borehole correlates well with the seismic data. Reduced-time
migration without the 3.5-ms time shift gives a more reliable
image than does conventional migration with the time shift.
The combination of SP transmission and PP reection migra-
tion almost completely illuminates the orebody boundaries.
Data description.An offset VSP experiment was carried
out by an anonymous seismic contractor in the Gulf of Mexico.
A plan view of the survey conguration is shown at the top of
Figure 14. The data consists of three 3-component gathers at
offsets of 150, 600, and 1500 m from the wellhead (Figure 14,
top). The 150-m offset gather consist of 82 receiver stations
from depths of 2650 to 3890 m at 15.3-m intervals. The 600 and
1500 m offset gathers contain 95 receiver stations each between
depths of 3050 and 4490 m, again at 15.3-m intervals. Sample
gathers are shown in Figure 15 with prominent modes labeled.
The upper gure shows the vertical component primarily con-
taining direct and reected P-waves. The horizontal component
(x) for the same shot offset is shown in the lower gure; this shot
gather records reections and transmissions from both P- and
SV-waves. A P-wave velocity model is generated from well
information (Figure 14, bottom left), and an S-wave velocity
model was estimated from near vertically incident PS trans-
mission and reection traveltimes (Figure 14, bottom right).

FIG. 14. Offshore VSP acquisition geometry. The top image shows a plan view of the relative location of the
sources to the wellhead, denoted by . Since the source was on a ship, there is some scatter in each offsets
source location. The bottom two gures show the velocity proles for P-waves (left) and S-waves (right). The
solid line represents the well log velocity function and the dashed line the smoothed velocity model used for
migration.
 1704 Sheley and Schuster
P-to-S velocity ratios of 1.6 and 2.7 was used for the salt and
sediment, respectively. Prior to migration, the data were reori-
ented by maximizing the P-wave energy (Ahmed, 1987). This
was rst done for the x- and y-components, maximizing the
energy on the x-component direct wave, then on the z- and x-
components, maximizing the z-component direct wave energy.
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
The events with the desired moveout velocity were picked and
attened. The attened gathers were median and bandpass l-
tered and unattened. Figure 16 shows the isolated reected
P-waves (top) and transmitted S-waves (bottom) for the 150-m
offset gather.
Migration results.For each of the three gathers described
in the previous section, the PP reections, PS reections, and
PS transmissions were migrated with reduced-time migration.
The grid spacing for the migration model was 3 m in both the
offset and depth directions. Conventional migration as well as
reduced-time migration were applied to these data, although
no signicant difference could be determined between the two.
We nd that the velocity models accurately describe the true
earth velocity. The average time difference between the calcu-
calc
lated direct-P traveltimes (sg ) and the picked direct-P trav-
obs
eltimes (sg ) was about 6 ms (six data time samples). Since
there is little difference between the two, we only show the
reduced-time examples.
We rst show the migration images for the 150-m offset
gather followed by the 600-m offset gather. For comparison,
FIG. 15. Shot gather for a source at 150 m with the
z-component at the top and the x-component at the bottom.
The y-component is not shown.
we generated a synthetic migration section (Figure 17, left)
based solely on the well log velocity.
Using Kirchhoff migration, we migrated the 150-m offset
z-component gather with isolated P-wave reections to obtain
the reection section shown in the center of Figure 17. The
prominent event at 3170 m is the base of the salt reection
and correlates with the same event in the synthetic section. A
similar event can be found somewhat higher in the reected PS
image (Figure 18, center). This gure shows another event at
2800 m which correlates with the top of salt. The transmitted
PS migration image (Figure 19, center) shows a strong event at
2800 m that correlates well with the top of salt. In this gure,
we reversed the polarity of the synthetic section to accentuate
the top boundary of the salt. Also in this gure, the reader may
notice that the events below the top of salt dip to the right. Since
the geologic interfaces in this region are assumed to be at and
appear to be so in the other migrated sections, we assume that
this dip may be caused by salt anisotropy.
Figures 17, 18, and 19, contain, respectively, the reected P,
the reected PS, and the transmitted PS migrated images for a
FIG. 16. Compare with Figure 15. Desired events were picked,
attened, median ltered, unattened, and bandpass ltered to
produce these gathers. f -k ltering was ineffective at isolating
the S-waves since the velocity of the S-waves in the salt is similar
to that of P-waves traveling in the clastic sediments. Also, some
of the S-waves events are spatially aliased in the sediments.
 Migration of transmission PS waves
source offset of 600 m. In Figure 18, notice the prominent events
at 3170 m, the base of salt, in the reected images. There is a
good correlation between the top of salt, above the uppermost
receiver at 3050 m, and the prominent event in the migrated
section. These results show that for VSP experiments, PS trans-
mission migration can be effective in illuminating boundaries
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
above the receiver array.
CONCLUSIONS
In this paper, we developed and tested two novel techniques
for crosswell and VSP seismic imaging: transmission PS migra-
tion and reduced-time migration. Transmission PS migration
complements PP reection migration by illuminating nearly
vertical boundaries between the source and receivers. In ad-
dition, reduced-time migration can decrease the imaging-time
error associated with either an incorrect migration velocity or
a static shift.
For a synthetic crosswell experiment, PS transmission ar-
rivals were migrated to locate a vertical boundary parallel to
the boreholes, a boundary that is invisible to PP reections. It
was also demonstrated that the reduced-time imaging condi-
tion could be applied to these data to decrease the focusing
errors due to an incorrect migration velocity.
Using the synthetic RVSP data, we applied PS and SP trans-
mission migration as well as PP reection migration to locate
FIG. 17. A synthetic migration section calculated solely from borehole velocities is shown on the left. There is
good correlation with the base of salt and the strong events in both the 150-m (center) and 600-m offset (right)
migrated gathers.
1705
the vertical boundary. Due to the chosen velocities, SP trans-
mission migration was able to image approximately two-thirds
of the boundary, PP reection migration one-half of the bound-
ary, and PS transmission migration one-third of the boundary.
For these synthetic data, the reduced-time imaging condition
migrated the energy more accurately than conventional migra-
tion for an incorrect P and S velocity model.
PS transmission migration of the Kidd Creek crosswell data
and the Gulf of Mexico VSP data demonstrated the feasibility
of imaging the boundaries of geologic bodies with large veloc-
ity contrasts. The use of reduced-time migration with the Kidd
Creek data clearly showed better focusing and more accurate
imaging than images obtained with conventional migration.
Applying reduced-time migration to the Kidd Creek data il-
lustrated that, in the presence of an unknown time shift and
migration velocity errors, transmitted SP arrivals and reected
PP arrivals can be used to accurately image the transmitting
and reecting boundaries. For these data, the base of the ore-
body would not have been imaged without the availability of
PS transmitted arrivals.
The Gulf of Mexico VSP data were successfully migrated
to reconstruct the top and bottom of the tabular salt body
even though the top of the salt body was above the receiver
array. This further illustrates the utility of migrating trans-
mitted PS arrivals in illuminating boundaries invisible to PP
reections.
 1706 Sheley and Schuster
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/
FIG. 18. There are strong events at both the base and top of the salt sheet for the 150-m offset source (center)
and at the base of salt for the 600-m offset source (right).
In summary, migration of transmitted converted waves and
reduced-time migration are two new tools that can be used
to reduce imaging error and illuminate nearly vertical bound-
aries, or dark parts of salt anks, orebodies, and other targets.
The reduced-time imaging condition also has the capability of
removing static errors due to source and receiver statics. The
main assumption of reduced-time migration is that the errors
encountered by the direct wave are representative of errors
in the entire migration space. If this assumption is violated,
reduced-time migration may give a worse result than conven-
tional migration. In the PP reection case, the direct wave and
the reected waves usually traverse different parts of the earth,
therefore improved imaging by reduced-time migration is ex-
pected to be the exception rather than the rule. The transmis-
sion wavepaths, in general, are subparallel to the direct wave
and therefore sample similar errors as the direct wave; hence
reduced-time migration is effective in transmission imaging.
REFERENCES
Ahmed, A., 1987, A new method for orienting the horizontal com-
ponents in three-component vertical seismic proling data: M.Sc.
thesis, Univ. of Utah.
Amundsen, L., Arntsen, B., and Mittet, R., 1993, Depth imaging of
offset vertical seismic prole data: Geophys. Prosp., 41, 10091032.
Balch, A. H., Chang, H., Hoand, G. S., Ranzinger, K. A. and Erdemir,
C., 1991, The use of forward- and back-scattered P-,S- and converted
waves in cross-borehole imaging: Geophys. Prosp., 39, 887914.
Chavez-P
erez,
S., 1997, Enhanced imaging of fault zones in south-
ern California from seismic reection studies: Ph.D. diss., Univ. of
Nevada, Reno.
Chavez-P
erez,
S., and Louie, J. N., 1996, Crustal imaging in southern
California using earthquake sequences: 7th Internat. Symp. on Deep
Seismic Proling of the Continents.
1998, Crustal imaging in southern California using earthquake
sequences: Tectonophysics, 286, 223236.
French, W. S., 1974, Two-dimensional and three-dimensional migra-
tion: Geophysics, 39, 265277.
Gardner, G. H. F., Ed., 1985, Migration of seismic data: Soc. Expl.
Geophys.
Gardner, G. H. F., French, W. S., and Matzuk, T., 1974, Elements of
migration and velocity analysis: Geophysics, 39, 811825.
Hu, J., 1999, Processing the Noranda Kidd Creek crosswell data: Uni-
versity of Utah Modeling and Tomography Project 1998 Annual
Report, 269288.
Hu, L., McMechan, G. A., and Harris, J. M., 1988, Acoustic prestack
migration of cross-hole data: Geophysics, 53, 10151023.
Langston, C. A., 1977. Corvallis, Oregon, crustal and upper mantle
receiver structure from teleseismic P and S waves: Bull. Seism. Soc.
Am., 67, 713724.
Meng, F., and McGaughey, J., 1996, Orebody hunting with crosswell
imaging methods: 66th Ann. Internat. Mtg., Soc. Expl. Geophys.,
Expanded Abstracts, 626629.
Payne, M., 1994, Considerations for high-resolution VSP imaging: The
Leading Edge, 13, 173180.
Qin, F., 1994, Application of traveltime modeling and imaging in cross-
well seismology: Ph.D. diss., Univ. of Utah.
 Migration of transmission PS waves 1707
Downloaded 03/26/17 to 212.26.2.132. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

FIG. 19. Migration of transmitted PS arrivals for 150-m (center) and 600-m (right) offset gathers. The traces of
the synthetic migration section have been rotated 180 from the previous two gures to emphasize the top of the
salt boundary. Notice the correlation between the top of salt in the synthetic and migrated gathers. The dip of
the lower events may be due to shear wave anisotropy within the salt.

APPENDIX A
RESOLUTION OF PS TRANSMISSION MIGRATION

By assuming a homogeneous velocity for both P- and S- Assuming that the trial image point is (0, 0) and the source and
waves, the traveltime ( ) for energy to propagate from a source receiver are diametrically opposed [i.e. (xr , zr ) = (xs , z s )],
at (xs , z s ) to a receiver at (xr , zr ) through an arbitrary image then we can rewrite equation (A-2) as
point at (x, z) is represented by
 d 2 xs 2 xs
1 =  +  . (A-3)
= (xs x)2 + (z s z)2 dx xs + z s
2 2 xs + z s2
2

 
1 Recognizing that xs2 + z s2 is the magnitude of the vector s,
+ (xr x)2 + (zr z)2 , (A-1)
equation (A-3) can be re-expressed as
where is the P-wave velocity and is the S-wave velocity.

d 2xs
To determine the resolution in the x-direction, we take the = 1 . (A-4)
dx |s|
derivative of the traveltime eld with respect to the horizontal
image coordinate x:
It can be seen that as approaches (i.e. the V p /Vs ratio
d 2 x xs nears one), the horizontal resolution (d/d x) goes to zero.
=  Conversely, as the V p /Vs ratio increases so does the horizon-
dx (xs x)2 + (z s z)2
tal resolution. A similar analysis may be performed for the
2 x xr vertical resolution to determine that vertical resolution is also
+  . (A-2)
(xr x)2 + (zr z)2 proportional the V p /Vs ratio.