Geophic J Int (1996)125, 30 38

Direct calculation of interval velocities and layer thicknesses from

seismic wide-angle reflection times

Kalachand Sain’ and K. L. Kaila2

I .Vc/rioiiu/ Geopkysical Reseurch Institute. Cppal Road, Hjderuhad 500 G07, Indin
’2- 16-106.Prushunririugcir. C~ppitlRocid. Hyderahad SO0 039, Indiu
t

Accepted 1995 October 7. Received 1995 October 4; in original form 1994 December 29

Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019

SUMMARY
Wide-angle seismic reflection traveltimes generated by large explosions are used
extensively to obtain both velocity and regional structure of the crust and upper
mantle. The method most commonly used to calculate interval velocities and thicknesses
(layer parameters) of a multilayered earth. supposed to consist of homogeneous,
isotropic and horizontal layers, is based on Dix’s ( 1955) interval velocity formula,
which needs a prior estimate of the rms velocity and traveltimes at zero offset. Use of
Dix’s hyperbolic approximation to the non-hyperbolic wide-angle reflection times
causes large errors in the determination of layer parameters. Currently, no direct
method exists to extract the velocity information from such wide-angle seismic data.
Here we propose a layer-stripping method, using the ray parameter, to calculate the
layer parameters directly in a vertically heterogeneous earth from a set of wide-angle
seismic reflection data. Synthetic reflection times contaminated by some realistic errors
have been used to demonstrate the efficiency and reliability of the algorithms. Field
examples using well-identified wide-angle reflection times illustrate the practical feasibil-
ity of the proposed method.
Key words: crustal structure, deep seismic reflection, seismic velocities.

of the zero-offset two-way time and the inverse square of the

1 INTRODUCTION
The reflection coefficients of seismic waves often increase with
offset, and reach a maximum value at the critical angle
(Richards 1961: Winterstein & Hanten 1985). The amplitudes
of the post-critical (wide angle) reflections are large due to
total internal reflection. Such reflections are usually prominent
on seismograms. even in a noisy area with a single fold
coverage. The recording of wide-angle seismic waves in
controlled-source seismology has become an important tool
for deciphering the variations in both seismic velocity and
crustal thickness in order to understand the tectonic framework
of an area. A review of such studies in central Europe, edited
by Giese, Stein & Prodehl (1976), contains many excellent
examples. For the study of deep crustal structure, extensive
use has been made of wide-angle reflections in the USSR, parts
of Europe, Canada, USA and India.
In a multilayered earth consisting of horizontal, homo-
geneous and isotropic layers, the reflection traveltime, T ( X ) ,
at an offset, X, is expressed by the Taner & Koehler (1969)
series as
T’(X)=C,+C,X’+C,X4+ ’.’.
The first two coefficients are respectively equal to the square
rim velocity. The other coefficients (C’, C,, . . .) are related to
layer parameters in a complicated way. The conventional
hyperbolic equation becomes inaccurate when the spread-
length-to-depth ratio and heterogeneity in the subsurface
increase. To improve the accuracy, some authors (Shah &
Levin 1973; Al-Chalabi 1974; May & Straley 1979; Hake,
Helbig & Mesdag 1984; Castle 1994; de Bazelaire & Viallix
1994; Kaila & Sain 1994; Thore, de Bazel & Ray 1994;
Tsvankin & Thomsen 1994) have suggested the use of higher-
order terms, but the gains d o not compensate for the dramatic
increase in computing time. In the past, the use of higher-
order terms has not received a wide application in the explo-
ration industry. The hyperbolic approximation (neglecting the
third- and higher-order terms of the reflection series), which is
valid only at small-offset ranges, is still used for seismic data
processing work.
Interval velocities and layer thicknesses from large-offset
reflection times have been calculated in various ways by
different authors. Dix’s (1955) interval velocity formula can be
used only when the rms velocity and two-way traveltime at
zero offset are known for a set of reflectors. As the offset
(1) distance increases, the estimated velocity, here called the slant-
path rms velocity (Robinson 1983), obtained by hyperbolic
fitting, is different from Dix’s rms velocity due to the cumulative

30 0 1996 RAS

effect of refractions of a single ray at each overlying interface
in a multilayered earth model. The percentage error in rms
velocity and zero-offset two-way traveltime, as determined
using the hyperbolic approximation to the increasing offset
reflection times without the control of offsets associated with
a low-angle region. becomes very large at the wide-angle ranges
(Kaila & Sain 1994). Al-Chalabi (1974) pointed out that the
use of the hyperbolic approximation to the large-offset reflec-
tion times causes large errors in the determination of interval
velocities and layer thicknesses unless an appropriate adjust-
ment is made. Velocity analysis from wide-angle reverse reflec-
tion times using the effective velocity, defined by Kaila &
Krishna (1979), has remained rather difficult.
For the calculation of interval velocities and thicknesses
from a series of large-offset reflection traveltimes, Sattlegger
(1965) proposed an iterative method that considers exact ray
paths in a multilayered earth in order to compute the forward
traveltimes, which are compared with the observed traveltimes
by a least-squares (Gauss--Newton)method. This is applicable
only to the small-offset traveltimes, which can be reasonably
approximated by a hyperbolic equation. For wide-angle reflec-
tion times, the solution obtained by the Gauss-Newton method
does not always converge, particularly when the initial model
does not lie very close to the true solution. Limond & Patriat
(1975) made use of Sattlegger's (1965) method in the compu-
tation of a forward response in order to match the wide-angle
reflection times in marine sonobuoy work by a least-squares
method in which only the velocity is varied while the zero-
offset traveltimes are constrained from the normal-incidence
reflection profile. Sain & Kaila (1994) show that, even without
constraints, layer parameters can be calculated accurately from
wide-angle reflection times by an iterative method. As com-
pared with the time-consuming iterative method, the present
method is direct and faster.
In order to calculate the interval velocity from large-offset
reflection times, many authors (Garmany, Orcutt & Parker
1979; Diebold & Stoffa 1981; Stoffa et al. 1981, 1982;
Schultz 1982) used the ~ - ptransformation, where p = ( d T / d X )
is the ray parameter and z equals T - p X , called the intercept
time. T is the two-way reflection time at an offset X . Our
proposed method does not need any transformation of ( X , T )
data. Direct methods (Gonzalez-Serrano & Claerbout 1984;
Nowroozi 1989, 1990) of calculating layer parameters do not,
of course, need any transformation, but they require the
recognition of an ( X , T ) pair with the same ray parameter at
the top and bottom of the layer concerned. For field data
containing many errors at a particular wide-angle recording
spread, recognition of the same ray parameter at the top and
bottom of a layer can be difficult. For a given ray parameter
associated with a particular wide-angle data point on reflec-
tions from the bottom of a layer, the corresponding data point
for reflection from the top of the layer may lie outside the
recording range. This is because wide-angle reflections are very
sensitive to ray parameters.
Wide-angle reflections and refractions are now used exten-
sively in crustal seismics to obtain the 2-D velocity structure
of the subsurface earth. This can be done either by ray-trace
forward modelling (McMechan & Mooney 1980; Cerveny &
Psencik 1984; Spence, Whittal & Clowes 1984; Zelt & Ellis
1988) or by inversion (Huang, Spence & Green 1986; Lutter,
Nowach & Braile 1990; Zelt & Smith 1992; Zelt & Forsyth
1994; Zelt et a!. 1994). For fast computation, a spatially non-
0 1996 RAS, G J I 125, 30- 38
Calculation of interval velocities 31
varying velocity structure is assumed in I-D forward traveltime
and amplitude modelling (Fuchs & Mueller 1971), Wiechert-
Herglotz inversion (Bullen & Bolt 1985) and T-p inversion
(Diebold & Stoffa 1981). Based on this assumption, we have
formulated a new method to calculate the interval velocities
and layer thicknesses directly from a series of wide-angle
reflection times.
2 THEORY O F THE PROPOSED METHOD:
VELOCITY A N A L Y S I S
We follow the conventional procedure of processing reflection
data in a layer-by-layer fashion working downwards. This
means that when we deal with the reflections from the nth
Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019
interface, all parameters for the overlying ( n - 1) layers are
known.
In a multilayered medium consisting of horizontal, homo-
geneous and isotropic layers, the offset X ( p ) for the nth
interface and corresponding reflection time T ( p ) are para-
metrically expressed (Taner & Koehler 1969) as
n
(3)
where the ray parameter p is defined by Snell's law as
sin Ok
p=- (4)
ck
0, is the angle of incidence at the kth layer. &, z k , L'k and xk
are defined in Fig. 1. X is the distance between the source and
the receiver corresponding to a ray parameter. The slope of a
straight line fitted to a segment of synthetic reflection data
( X , T ) produces the ray parameter quite accurately for the
centre of the segment, a point which is here referred to as the
mid-distance of the segment (Xmd).The determination of the
ray parameter remains accurate even if the synthetic data are
contaminated by any percentage of realistic errors. The exact
ray parameters at various offsets for the nth interface of a
given velocity model can be calculated using eq. (4)correspond-
ing to different prescribed take-off angles at the source.
For the velocity model tabulated in Fig. 2(a), the true ray
parameters from the bottom interface of the model are
shown in Figs 2(a) and (b) against the offset-to-depth ratio.
Corresponding to various ray parameters, the reflection times
at different offsets are calculated using eqs (2) and (3) respect-
ively. Here the true model has six layers. In this case, the
various take-off angles lie below a maximum value, which is
defined as
where L ' ~and ch are the velocities of the first and the greatest
velocity layers in the model. To simulate field situations, we
add to each traveltime some percentage of itself in order to
create realistic noise levels. We now show how accurately we
can calculate ray parameters from any increasing offset reflec-
tion times. For this, we segment the calculated traveltimes into
groups containing, say, 50 offsets, an arbitrary number of
datapoints (as large a data set as possible should be used for
noisy data), such that the first segment includes the first to
32 K . Sail? arid K . L. Kuilu

SOURCE RECEIVER
DISTANCE
r\ VI ZI 1

Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019

/-

nth interface

I (n+l)th layer

Figure 1. Reflection ray path in a homogeneous and isotropic medium consisting of horizontal layers, with various parameters defined.

-
- Jdwty TMrnur
m
010
uu)
2.0
3.0
Du)
l.5
2.0
c 0.101 //
w
I: 40 2s
a 54 3.0
a 60 loo
Q 0 0 5
7.0 6.0 True ray parameter
a
>
a ..
- True ray parameter
Calculated ray parameter
from noise f r e e traveltimes
>
0.,
Calculated ray parameter
from 1% noisy traveltimes
1 1 - J
000 I I 1

0 1 2 3 4 5 6 0 1 2 3 4 5 6
OFFSET /DEPTH OFFSET/DEPTH
(cl (d)
2 3 4 5 1 2 3 4 5 6
0 _ 1_ _ _ t r u e thicknhss
I T l O
- -
UI True velocity I
Calculated thickness \ Calculated
m ,

x
\
E -E
x u
--
0 0 0 Calculated velocit
n 0 YJ.

->- 10
E
!
10
b U
0
U
0 A
J w
W >
>
20 L 120 20
Figure 2. A comparison between true ray parameters and those determined by the straight-line for various segments with increasing offset
reflection times is shown for (a) noise-free data and ( b ) noisy synthetic data calculated from the bottom interface of the model tabulated in (a). A
comparison between true interval velocity and thickness of the sixth layer and those calculated by the present method is shown for (c) noise-free
data and (d) noisy data. The comparison is shown with increasing offset reflection times.

0 1996 RAS, GJZ 125, 30-38


fiftieth data points, the second segment contains the second to
fifty-first, and so on until all data points are included. We
calculate the ray parameter through a straight-line fit to the
traveltimes of each segment and assign it to the mid-distance
of the segment. These ray parameters have been plotted in
Figs 2(a) and (b), respectively, against the offsetldepth ratio
for noise-free and 1 per cent noisy synthetic data. We have
seen that any type of error can be added, and the straight-line
fit gives similar values for the ray parameter. It should be
noted that a 1 per cent error for the reflection time for the last
interface of the model includes a maximum value lying between
100 and f300 ms, which is rather high in comparison to
the noise encountered in the field. It is evident from Fig. 2 that
this method allows the ray parameters to be calculated quite
accurately. Since wide-angle reflections are observed beyond
the offset-to-depth ratio of 2, we show the calculation up to
an offset-to-depth ratio of 6, within which all wide-angle
reflections may be recorded on the seismograms. Of course,
the ray parameter can also be calculated from further offset
traveltimes.
For the calculation of interval velocity and thickness of the
nth layer, we calculate the traveltime, Tmd,at the mid-distance,
Xmd, of each segment by smoothing the available traveltimes
of the segment through a local fit using a hyperbolic equation
such as:
Tid = COf + ClfXkd > (6)
where Cof and CIf are the fitted coefficients. For downward
processing of reflection times from the nth interface, we know
all layer parameters for the top (n - 1 ) layers. Referring to
Fig. 1 and using eqs (2) and (3) we can write
where K, and K , can be computed from the layer parameters
of all overlying layers. Eqs (7) and (8), representing time and
distance for the nth layer, differ from eqs (12) and (13) given
by Nowroozi (1989). The time and distance in Nowroozi's
( 1989) paper are the respective differences in traveltimes and
offset distances for the reflected rays from the top and bottom
of the layer in which rays start at the source with the same
take-off angle or ray parameter. Such recognition of the ray
parameter is not needed in this work. The present method
works for any ray parameter, using a layer-stripping method.
Now, from eqs (7) and (8) we have
(9)
Using eq. (9), the velocity, G ~ and the thickness, z,, of the nth
, a=- 2 '
layer can be directly calculated as
Figs 2(c) and (d) show the calculated layer parameters for the
sixth layer from various segments of noise-free and noisy
synthetic data, respectively. For clarity, layer parameters have
0 1996 RAS, G J I 125, 30-38
Calculation of intercul celocities 33
been shown for every fifth segment. This demonstrates the
accuracy of the present method for calculating interval velocity
and layer thickness from large-offset or wide-angle rellection
times. Eqs (10) and (11) indicate the velocity-depth (or thick-
ness) ambiguity problem (Bickel 1990; Lines 1993; Ross 1994).
which is inherent in velocity estimations for the case of
vertically propagating waves ( p = 0).
3 MEANERRORS
In reality, the seismic velocity in a given layer varies both
laterally and vertically, and the interfaces undulate. Therefore.
the assumption of the flat-layered earth model produces aver-
aged interval velocities and thicknesses of various layers.
Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019
Hence, each layer parameter is associated with some mean
error which can be calculated for the nth layer as follows.
Because the interval velocities and thicknesses of all I I layers
are known, the ray parameters at various offsets of the
reflection data from the nth interface can be calculated using
eq. (3), which is based on Sattlegger's (1965) algorithm in a
modified form. In this scheme, we first calculate the maximum
take-off angle, Om,,, for the above n-layered velocity model
using eq. ( 5 ) . A take-off angle equal to half of ,H, (or the
average between the maximum take-off angle and the zero
take-off angle) is provided at the source, and the ray parameter
is calculated using eq. (4). Then we calculate the offsel for the
ray reflected from the nth layer as per eq. (3). If the calculated
offset is more (or less) than the required offset, a take-off angle
equal to the average of the previous take-off angle and the
zero (or maximum) take-off angle is determined and the offset
is again calculated using eq. (3). Next, we decide on a value
for the take-off angle by taking the average of the previous
take-off angles for which offsets lie immediately on either side
of the required offset. Thus the take-off angle or ray parameter
is searched for iteratively until the reflected ray reaches the
required offset within a given tolerance. The process is repeated
for all offsets, and the calculated ray parameters are then used
in eqs (2) and (3) for the computation of traveltime and offset
contributions from all ( n - 1) overlying layers. Subtraction of
these traveltimes and offsets from the respective traveltimes
and offset distances coming out of the nth interface produces
the traveltimes and offsets associated only with the nth layer.
The subtracted traveltimes, AT, and offsets, AX,,, are related
to the interval velocity, c, and the thickness, z,, of the nth
layer through a hyperbolic equation:
AX: 42:
AT: =7 +7 = aAXf + b ,
11, c,
where
1
(13)
2'"
and
The errors in observed (X, T ) data and the assumption of a
flat-layered model will produce mean errors both in the interval
velocity and the thickness of the nth layer. The mean errors
6(u) of u and h(b) of b as illustrated by Topping (1963) are
34 K . Sain and K . L. Kaila

given by r K 71'2

where rn(ATi) is the mean error of squared offset times between
the subtracted traveltimes, A X i , and the traveltimes of the nth
4 SYNTHETIC A N D F I E L D EXAMPLES
Before any given method is applied to the field data, it should
first be tested on synthetic data where the model parameters
are known. A simplified six-layer velocity model of the conti-
nental crust is shown in Fig. 3(d) by the solid curve. The true
velocities are 2.10,3.00,4.00, 5.00,6.00 and 7.00 km s f l respect-
ively, from the first to the sixth layers. The corresponding true

Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019

layer only, calculated using the hyperbolic relation between thicknesses are 1.50, 2.00, 2.50, 3.00, 10.00 and 15.00km
the velocity, u,, and thickness, zn, at various offsets, AXni. K is respectively. Some wide-angle reflection times are analytically
the total number of ( A X n i , A T i ) data points from the nth computed using the parametric equation (2) for offsets defined
interface. Now, from eqs (13) and (15), and from (14) and by eq. (3) by providing different take-off angles at the source
(16), the mean error in the velocity, b(u,), and the mean error corresponding to the various interfaces. 1 per cent random
in the thickness of the nth layer, S(z,), can be calculated as errors with Gaussian distributions are then added to these
reflection times to simulate a field situation. The synthetic data
r limited to the wide-angle range are shown in Figs 3(a), (b)
and (c) for the first and second, third and fourth, and fifth and

.. 1% noisy reflection traveltimes

Computed traveltimes of the
d estimated velocity model
= 2.01
LLJ I I 1 I 1
10.0 20.0
_ . 30.0 40.0 50.0 60
DISTANCE (km 1 DISTANCE (km 1

w
-: 7.0
I
t-
w
A 6.0
2 5.0
Oz
- True model
t- - -- Calculated model
d 4.0 t
-

= 3.01
LJ

40.0 60 0
I I
80.0
I
100.0
I
120.0
40 L

DISTANCE (km 1
Figure 3. Synthetic reflection times with added noise, which are computed for various interfaces of the true model shown in (d), are compared
with the traveltime curves generated for the estimated model in (a), (b) and (c) for the first and second, third and fourth and, fifth and sixth layers,
respectively. The six-layer true velocity model is compared in (d) with the velocity model obtained by the present method.

.. - RAS, GJI 125, 3&38

0 1996

sixth layers, respectively. The random errors added to the first
to sixth reflected phases are in the ranges 240, k75, 2 100,
+- 120, 5 150 and k200 ms, respectively. A reduced time scale
with a reduction velocity of 6.0 km s - l has been used through-
out the paper to present all traveltimes in a compact form. By
applying the proposed method to these synthetic data in a
layer-by-layer fashion, we obtain the velocity model shown by
the dashed line in Fig. 3(d). The interval velocities associated
with mean errors, calculated as per eqs (10) and (17) for the
first to sixth layers, are 2.07F0.02, 3.03 50.04, 4.03 AO.07,
5.06k0.11, 6.05k0.17 and 7.06f0.19kms-', respectively.
The corresponding thicknesses and associated mean errors,
calculated as per eqs (11) and (181, are 1.51 k 0.05, 2.03 +_ 0.18,
2.53k0.35, 3.08f0.48, 10.15f0.55, and 14.90i0.62km.
The rms residual between the synthetic data and the comput-
ed traveltimes of the calculated velocity model are 0.03, 0.05,
0.06, 0.06, 0.08 and 0.11 s respectively. The traveltimes shown
in Figs 3(a), (b) and (c) indicate a very good match between
the synthetic data and the computed traveltimes of the calcu-
lated model. It is also evident that the calculated velocity
model agrees quite well with the true model.
We now use two field data sets to demonstrate the practical
feasibility of the method. Since we have assumed a 1-D model,
the method produces averaged interval velocity and thickness
of various layers. Fig. 4 shows a trace-normalized record
section on a reduced time scale corresponding to shot point
235 along the Hirapur-Mandla deep seismic sounding profile
(Kaila et al. 1987, 1989). The traveltimes of first arrivals
(marked by P , and PB phases) observed in the record section
are plotted in Fig. 5(a) on a reduced time scale. The analysis
of the first-arrival traveltime data using the method of Kaila
& Narain (1970) produces a very thin (0.83 +0.02km) sedi-
mentary layer of velocity 5.09 _+ 0.01 km s- overlying the
crystalline basement of velocity 5.97 k 0.03 km s- '. The rms
residuals between the data being fitted and the fitted lines for
the above two refraction line segments are 0.016 and 0.032 s.
Two intracrustal reflections, marked by P1 and P z , have been
iden$&d by comparing seismograms from the reciprocal shot
points along the profile. Some 'floating reflections' appear on
the seismogram (Fig. 4), but they are not considered as we did
DISTANCE (KM)
Figure 4. A trace-normalized field seismogram (plotted at a reduced time scale) corresponding to shot point 235 along the Hirapur-Mandla DSS
profile, central India (Kaila et al. 1987, 1989), showing refraction phases marked by Ps and P,, and wide-angle reflection phases marked by P',
P2and P'.
0 1996 RAS, GJI 125, 30-38
Calculation of interval velocities 35
not observe a n y similar reflected phases ,on seismograms for
reciprocal shot points. Phases have been identified based on
the reciprocity test (Zelt & Forsyth 1994) of the picked
traveltime data from the reciprocal shot points. The P' reflected
phase is regarded as wide-angle reflections from the bottom of
the crystalline basement, as it extends asymptotically to the
basement first arrivals. The strong reflected phase marked by
P M represents the wide-angle reflections from the Moho bound-
ary, and this phase has been identified from the reciprocal shot
point (SP 80 along the profile). The downward extension of
the basement is obtained from the traveltimes of the P' phase
using the present method. This gives an interval velocity of
5.93 0.02 km s-I and thickness of 7.05 k 0.04 km for the
second layer with an rms residual of 0.029 s. The velocity of
Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019
the second layer is taken as the average (5.95 km s - * ) of the
velocities obtained for the layer from both the first-arrival
and the reflection data. The second reflected phase (P2)
produces an interval velocity of 6.65 k 0.02 km s- ', corre-
sponding to a thickness of 15.18 +_ 0.03 km and an rms residual
of 0.026 s. The velocity of 6.98 5 0.03 km s-' and the thickness
of 21.69 If: 0.21 km calculated from the traveltimes of the wide-
angle Moho reflections (PM) are the averaged interval velocity
and thickness of the lower crust in the Vindhyan basin, India.
The present study reveals the Moho depth of 44.75 km, which
is comparable to the reported Moho depth (42-46 km) in this
region (Kaila et a/. 1987, 1989). The velocity structure and the
Moho depth are also comparable to the worldwide average
velocity structure and the thickness (41.4 If: 6.2 km) of the crust
in the shield regions (Christensen & Mooney 1995). The
traveltime residual corresponding to the P M phase is 0.05 s.
Fig. 5(b) shows the estimated velocity model.
Kaila et aI. (1990) have identified two wide-angle reflected
phases in a field seismogram (marked by p5.* and p6.6 in Fig. 7
of their work) from layers below the basement, corresponding
to shot point 25 along the Sadra-Mehmadabad-Degam part
of a long DSS profile in the North Cambay and Sanchor
basins, India. The traveltimes of these two phases are shown
in Fig. 6(a). Since we process the data in a layer-by-layer
fashion, we need velocity information from the surface down
to the basement. As the 2-D velocity model at shallow levels
60 40 20
36 K . Sain and K . L. Kaila
0
*..-*
~
-
~
Reflection traveltimes
Refraction
1 ~ traveltimes
1 -
Y

1.1
-Traveltime
calculated model ..
curves o f
v)
Y

W
5
3
W
Bu
t-
d
W
Q:

Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019

VI = 5.05 KM/S v2=5 . 9 5 KM/S
I 1

i4
A

V3= 6.65 KM/S

P
15 Y
Y

I
L
r
c
L I I

(obtained by forward modelling of seismic refraction data) in

this region (Kaila et ul. 1990) is more or less flat, we have
averaged out the velocity structure in the first three layers to
a simplified three-layer 1-D velocity model, indicated by veloc-
ity values (in km s-') at the right-hand side and thicknesses
(in km) by dashed lines in Fig. 6(b). The first layer, with a
velocity of 2.00 km s f 1 and thickness of 1.30 km, the second
layer, with a velocity of 3.30 km s-' and thickness of 3.90 km,
and the third layer, with a velocity of 4.80 km s-' and thickness
of 1.50 km, are fixed while calculating the layer parameters of
deeper layers. The first wide-angle reflected phase produces an
interval velocity of 6.38 0.04 km s-' and a thickness of
3.57 2 0.06 km with an rms residual of 0.031 s. The second
reflected phase produces an interval velocity of 6.04f
0.03 km s f 1 and a thickness of 2.21 kO.08 km with an rms
residual of 0.028 s. The estimated layer parameters for the
fourth and fifth layers are shown by velocity values in km s-'
at the right-hand side of Fig. 6(b), and the thickness or depth,
by dashed lines. The theoretical reflection times corresponding
to the estimated velocity model are shown by the curves in
Fig. 6(a). As the present method is based on ray theory and
the calculation of ray parameters at the mid-distance of a data
set is quite accurate, there is no problem even with the
estimation of a low-velocity zone (LVZ), as has been calculated
for the second reflected phase in the upper crust. The 2-D
velocity model (Kaila et al. 1990) has been superimposed on
the calculated 1-D velocity model in Fig. 6(b) to show the
comparison and reliability of the proposed method.
5 CONCLUSIONS
The proposed method of calculating interval velocities and
layer thicknesses is applicable to a set of reflection times at
any non-zero offset distance. The method is very simple and
direct in nature, and can be easily implemented on a computer
for a quick estimation of both velocity and structure. The 1-D
velocity models, obtained by the present method from different
sets of wide-angle reflection times corresponding to various
shot points along a deep seismic reflection profile, will produce
a pseudo 2-D velocity picture of the sub-surface when inte-
grated together, and can provide a good starting model for
further refinement of the velocity structure by 2-D modelling
(Bishop et a/. 1985; Bording et al. 1987; Lutter et a/. 1990 Zelt
& Smith 1992).
ACKNOWLEDGMENTS
The authors are grateful to Dr H. K. Gupta, Director NGRI,
for his constant encouragement to carry out this work and
kind consent to publish the same. They also express their
gratitude to Dr C. A. Zelt for his constructive comments and
useful suggestions to improve the quality of the work, and to
Dr L. R. Lines for critically reading the manuscript. Finally,
the first author (KS) wishes to thank the Council of Scientific
and Industrial Research, India, for providing him with a
research fellowship.

0 1996 RAS, GJI 125, 30-38

 Calculation of interval velocities 31

60 1987. Application of seismic travel time tomography, Geophys. J . R.

astr. Soc.. 90,285-303.

-- 5.0
w
m
Bullen, K.E. & Bolt, B.A., 1985. An introduction to the theory of
seismology, Cambridge University Press, Cambridge, UK.
Castle, R.C., 1994. A theory of normal move out, Geophysics, 59,
E
t-
983-999.
Cerveny, V. & Psencik, I., 1984. PROGRAM s~is83---Numericalmodeling
A
of seismic wave fields in 2-D laterally varying layered structure by
4.0 the ray method, in Documentation of earthquake algorithms, ed.
4
a
b- t .... Observed traveltime d a t a Engdahl, E.R., World Data Center (A) for Solid Earth Geophys.,
Rep. SE-15.
o 3.0 Traveltime curves of Christensen, N.I. & Mooney, W.D., 1995. Seismic velocity structure
w and composition of the continental crust: a review, J . geophys. Res.,
LT calculated model
100, B7, 9761-9788.
de Bazelaire, E. & Viallix, J.R., 1994. Normal move out in focus,
2.0 Geophys. Prospect., 42, 477-499.

Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019

0.0 10.0 20.0 30.0 Diebold, J.B. & Stoffa, P.L., 1981. The travel time equation r-p
DISTANCE (km) mapping and inversion of common midpoint data. Geophysics,
46,238 -254.
Dix, C.H., 1955. Seismic velocities from surface measurements,
Geophysics, 20, 68-86.
Fuchs, K. & Mueller, G., 1971. Computation of synthetic seismograms
25 27 29 with the reflectivity method and comparison with observations,
I
110.0 rn 30.0 Geophys. J . R . astr. SOC.,23, 417-433.
Garmany, J., Orcutt, J. & Parker, R., 1979. Travel time inversion: A
geometrical approach, J. geophys. Res., 84, 361 5-3622.
I 3.3 3.30 I Giese, P., Stein, A. & Prodehl, C., 1976. Explosion seismology in central
Europe-data and results, Springer Verlag, Heidelberg.
Gonzalez-Serrano, A. & Claerbout, J.F., 1984. Wave equation velocity
analysis, Geophysics, 49, 1432-1456.
Hake, H., Helbig, K. & Mesdag, C.S., 1984. Three term Taylor series
for t2--xZcurves over layered transversely isotropic ground, Geophys.
Prospect.. 32, 828-850.
Huang, H., Spence, C. & Green, A,, 1986. A method for the inversion
of refraction and reflection traveltimes for laterally varying velocity
structures, Bull. seism. Soc. Am., 76, 837-846.

15
6.6
J Kaila, K.L. & Krishna, V.G., 1979. A new computerized method for
:2 Shot point Location with number
2.0; 1-0 model
-
5.6 2-D model
Figure 6. Traveltimes of two wide-angle reflected phases (marked by
P5.' and P6.6 in Fig. 7 of Kaila et al. 1990) from layers below the
basement and their comparison with the traveltimes generated for the
estimated velocity model. (b) The estimated 1-D model is shown
superimposed on the available 2-D velocity structure (Kaila et al.
1990) for which thickness or depth is indicated by a solid line and
velocity values (km s- ') are shown in the middle of the model. The
velocities of the 1-D model are shown in italics and the thicknesses by
the dashed lines. The first three layers of the Kaila et al. (1990) model
have been averaged into a 1-D model, and the fourth and fifth layer
parameters have been calculated by the present method.
REFERENCES
Al-Chalabi, M., 1974. An analysis of stacking, average and interval
velocities of a horizontally layered ground, Geophys. Prospect.,
22, 458-475.
Bickel, S.H., 1990. Velocity4epth ambiguity of reflection travel times,
Geophysics, 55, 266-276.
Bishop, T.N., Bube, K.A., Cutler, R.T., Love, P.L., Resnik, J.R., Shuey,
R.T., Splinder, D.A. & Wyld, H.W., 1985. Tomographic determi-
nation of velocity and depth in laterally varying media, Geophysics,
50,903-923.
Bording, R.P., Gersaztenkorn, A,, Lines, L.R., Scales, J.A. & Treitel, S . ,
finding effective velocity from reversed reflection travel time data,
Geophysics, 44, 1064-1076.
Kaila, K.L. & Narain, H., 1970. Interpretation of seismic refraction
data and the solution of the hidden layer problem, Geophysics,
35, 613-623.
Kaika, K.L. & Sain, K., 1994. Errors in RMS velocity and zero offset
two-way time as determined from wide-angle seismic reflection
travel times using truncated series, J. seism. Explo., 3, 173-188.
Kaila, K.L., Murty, P.R.K., Mall, D.M., Dixit, M.M. & Sarkar, D.,
1987. Deep seismic soundings along Hirapur-Mandla profile,
Central India, Geophys. J . R. astr. Soc., 89, 399-404.
Kaila, K.L., Murty, P.R.K. & Mall, D.M., 1989. The evolution of the
Vindhyan basin vis-a-vis the Narmada-Son lineament, Central India,
from deep seismic soundings, Tectonophysics, 162, 277-289.
Kaila, K.L., Tewari, H.C., Krishna, V.G., Dixit, M.M., Sarkar, D. &
Reddy, M.S., 1990. Deep seismic sounding studies in the north
Cambay and Sanchor basins, India, Geophys. J . Znt., 103, 621-637.
Limond, W.Q. & Patriat, Ph., 1975. The accuracy of determination of
seismic velocities from variable angle reflection (disposable sono-
buoy) records, Geophys. J. R. astr. Soc., 43, 905-938.
Lines, L., 1993. Ambiguity in analysis ofvelocity and depth, Geophysics,
58, 596-597.
Lutter, W.J., Nowach, R.L. & Braile, L.W., 1990. Seismic imaging of
upper crustal structure using travel times from the PASSCAL Ouachita
experiment, J . geophys. Res., 95, 4621-4631.
McMechan, G.A. & Mooney, W.D., 1980. Asymptotic ray theory and
synthetic seismograms for laterally varying structures: theory and
application to the Imperial valley, California, Bull. seism. Soc. Am.,
70, 2021-2035.
May, B.T. & Straley, D.K., 1979. Higher-order move out spectra,
Geophysics, 44, 1193-1207.

0 1996 RAS, GJZ 125, 30-38

38 K . Sain and K . L. Kaita
Nowroozi, A.A., 1989. Generalized form of the Dix equation for
calculation of interval velocities and layer thicknesses, Geophysics,
58, 596-597.
Nowroozi, A.A., 1990. Interpretation of seismic reflection records:
Direct calculation of interval velocities and layer thicknesses from
travel times, Pure appl. Geophys., 133, 103-115.
Richards, T.C., 1961. Motion of the ground on arrival of reflected
longitudinal and transverse waves at wide-angle reflection distance,
Geophysics, 26, 277-297.
Robinson, E.A., 1983. Seismic velocity analysis and the convolution
model, pp. 71-90, Reidel, Dordrecht.
Ross, S.H., 1994. The velocitydepth ambiguity in seismic reflection
data, Geophysics, 59, 830--843.
Sain, K. & Kaila, K.L., 1994. Inversion of wide-angle seismic reflection
times with damped least squares, Geophysics, 59, 1735-1744.
Sattlegger, J., 1965. A method of computing interval velocities from
expanding spread data in the case of arbitrarily long spreads and
arbitrarily dipping interfaces, Geophys. Prospect., 13, 306-318.
Schultz, P.S., 1982. A method of direct estimation of interval velocities,
Geophysics, 47, 1657- 1671.
Shah, P.M. & Levin, F.K., 1973. Gross properties of time-distance
curves, Geophysics, 38, 643-656.
Spence, G.D., Whittal, K.P. & Clowes, R.M., 1984. Practical synthetic
seismograms for laterally varying media calculated by asymptotic
ray theory, Bull. seism. SOC.Am., 74, 1209-1223.
Stoffa, P.L., Buhl, P., Diebold, J.B. & Wenzel, F., 1981. Direct mapping
of seismic data to the domain of intercept time and ray parameter:
A plane wave decomposition, Geophysics, 46, 255-267.
Stoffa, P.L., Diebold, J.B. & Buhl, P., 1982. Velocity analysis for wide
aperture seismic data, Geophys. Prospect., 30, 25-57.
Taner, M.T. & Koehler, F., 1969. Velocity spectra digital computer
derivation and application of velocity functions, Geophysics, 34,
859-881.
Thore, P., de Bazelaire, E. & Ray, M.P., 1994. The three parameter
equation: an efficient tool to enhance the stack, Geophysics, 59,
297 -308.
Topping, J., 1963. Errors of observation and their treatment,
pp. 105-109, Chapman and Hall Ltd, London.
Tsvankin, I. & Thomsen L., 1994. Non-hyperbolic reflection move out
in anisotropic media, Geophysics, 59, 1290- 1304.
Winterstein, D.F. & Hanten, J.B., 1985. Super-critical reflections
observed in P and S wave data, Geophysics, 50, 185195.
Zelt, C.A. & Ellis, R.M., 1988. Practical and efficient ray tracing in
two-dimensional media for rapid traveltime and amplitude forward
Downloaded from https://academic.oup.com/gji/article-abstract/125/1/30/703562 by guest on 08 May 2019
modeling, Can. J. Expl. Geophysics, 24, 16-31.
Zelt, C.A. & Forsyth, D.A., 1994. Modeling wide-angle seismic data
for crustal structure, southeastern Grenville Province. J . geophys.
Res., 99, 11 687-1 1 704.
Zelt, C.A. & Smith, R.B., 1992. Seismic traveltime inversion for 2-D
crustal velocity structure, Geophys. J. Int., 108, 16-34.
Zelt, C.A., Forsyth, D.A., Milkereit, B., White, D.J., Asudeh, I. & Easton
R.M., 1994. Seismic structure of the Central Metasedimentary Belt,
southern Grenville Province, Can. J . Earth Sci., 31, 243--254.
0 1996 RAS, G J i 125, 30-38