Demultiple

DP START Oslo
2004
 At the end of this Session, you will be able to:
• Define the terms primary energy and multiple energy as used in data
processing.
• Define the terms multiple order and period as used in processing.
• Define the following water bottom multiple, free surface multiple, peg-leg
multiple and internal multiple
• Produce a diagram that shows the travel path of each of the following
multiple types: water bottom multiple, free surface multiple, peg-leg
multiple, internal multiple

DP START Oslo
2004
 theory2.ppt

Primary and Multiple energy

• Primary energy is energy which has been reflected only once, and so is
a true image of the reflector from which it arose

• Multiples are events reflected more than once

• Multiples hinder interpretation
•False events
•Incorrect amplitudes
• Multiples can be very difficult……
•to identify
•to remove

DP START Oslo
2004
General Properties of Multiples

• Low velocity (high moveout)

– Velocity increases with depth

• High amplitude
– Velocity increases with depth, less geometric spreading

• Periodic
– Repeated cycles in horizontal layers

• Predictable
– From primaries

DP START Oslo
2004
 theory2.ppt

Multiple Order

• This is the number of additional bounces that a seismic wave has

undergone in addition to the original primary reflection.

• Therefore, a second-order multiple is a signal that has undergone a

primary reflection plus two additional bounces off the same layer.
P 1st 2nd

DP START Oslo
2004
 theory2.ppt

Multiple Period

• This is the time between each successive bounce of the multiple

series.

•Multiple periods are generally classified into three types:

• 1.Short
• 2.Intermediate
• 3.Long

DP START Oslo
2004
 theory2.ppt

Multiple Period

•Short period multiples generally have a period of

100 or 200 msec. They can usually be removed
using short operator deconvolution.

•Long period multiples generally have a period of

1/2 second or longer. They can usually be removed
using long operator deconvolution.

•Intermediate period multiples fall directly between

these two types. Also, they are generally
characterised by the fact that they are rarely
removed by any deconvolution.

DP START Oslo
2004
 Order of Multiples
Event recorded, dependent on receiver
position:
A - Primary
B - First order multiple primary
C - Second order multiple

A B C

multiple
1st order period
multiple

2nd order
multiple

DP START Oslo
2004
 Water Bottom Multiple
• An extremely common source of multiple reflection
occurs in marine seismic work because of the presence of
the water layer.

• This is because both the water/air interface and the water

bottom are characterized by having a large change in
acoustic impedance.

• These large reflection coefficients cause a high

proportion of the seismic energy to be trapped in the water
Water-bottom
layer, and allows little signal to be transmitted into the multiple
earth.

DP START Oslo
2004
 Water Bottom Multiple

The reflection coefficient of the water/air

interface approximates to -1, so as can be seen,
the polarity of the signal changes after each
reflection from this interface (ie. there is a 180°
phase change at each reflection).

DP START Oslo
2004
 Pegleg Multiple

• Just as the water layer acts as a direct

source of strong multiples, so may it also
act as a source of multiple reflections for
primary reflections which arise within the
earth itself.

• In this case the multiples that are

generated are referred to as peg-leg
multiples, because of their characteristic
ray paths.
Water-bottom Pegleg
pegleg multiple Multiple
• Their paths are asymmetric.

DP START Oslo
2004
 Free Surface Multiple

• An impedance contrast occurs at the land/air

boundary in land acquisition just as at the water/air
boundary in marine acquisition.

• Again as in marine case the reflection coefficient of

the air/land surface interface approximates to -1, so
the polarity of the signal changes after each
reflection from this interface (ie. there is a 180°
phase change at each reflection).

Free-surface
multiple

DP START Oslo
2004
 Internal Multiple

• The only prerequisite for

multiples to occur is that
there should be at least two
strong reflectors.

• Therefore, multiples may

arise within the earth at any
interface, provided that the
two reflectors have "large"
reflection coefficients
Inter-bed or
internal multiple

DP START Oslo
2004
 At the end of this Session, you will be able to:

• Describe how to distinguish primaries from multiple events using their

velocity.
• Describe the removal of multiples using stack.
• Describe the removal of multiples using fk demultiple.
• Describe the transforming of events into the tau-p domain.
• Describe the removal of mutliples using tau-p demultiple.
• Define the following terms PRT, p traces, moveout range and mute range.
• Describe the removal of multiples using wave equation demultiple.

DP START Oslo
2004
 theory2.ppt

Removal of multiples

The removal of multiples generally relies on either or both of two

recognisable characteristics.

•Multiples will generally go on and on,

repeating with the same time interval and
gradually decreasing in amplitude (think of
repeated "bounces" of energy in a water
layer).
• Multiples will generally appear on our
CDP gathers with a velocity slower than
the primary velocity. They will have larger
moveout values (more curvature).
DP START Oslo
2004
 theory2.ppt

Primary and Multiple Velocity

• As the primary and multiple energy has both travelled through the
same layer the multiple just spent longer in the layer then what’s their
velocity relationship.

They have the same velocity

P 1st 2nd

DP START Oslo
2004
 Velocity
Constant Velocity Stack (CVS)
So looking at our constant velocity stack
we can recognise a multiple as having the
same velocity as the primary.

We will see the same thing on our primary

semblance display where multiples will multiple
appear directly below primaries. period

But as we know velocity should increase 1st order

the further into the earth you go so multiple
mutliples will have a slower velocity than
the events surrounding them.
2nd order
As we have seen they will appear to be multiple
under corrected on your cmp gathers.

DP START Oslo
2004
 Stack

CMP stack of nmo corrected

gathers will result in
enhancement of the primary
energy and degradation of
the multiple energy.

CMP gather and stack after NMO

correction using primary velocity
DP START Oslo
2004 function
Stack

Easy
Cheap
Improves S/N

 Incomplete multiple suppression

 Output is poststack - limited AVO available

DP START Oslo
2004
 Near Trace Mute

Removing the near traces from

our cmp gather results in further
degradation of the multiples.
As the multiple event appears
flattest on the near traces and
so stacks up best for the near
traces.

CMP gather and stack after NMO

correction using primary velocity
DP START Oslo
2004 function
 FK demultiple

1. The normal procedure is to

overcorrect the primaries
while leaving the multiples
primary
undercorrected.

2. Transform to FK domain.
multiple

3. Remove all events at K > 0

i.e. all multiple energy muted.
primary

multiple

CMP gathers after NMO correction using

multiple velocity function
DP START Oslo
2004
 FK demultiple

CMP gather x NMO correction x F

f

Multiple
Mute
Primary
NMO FT

t t
Using velocity between - K k +
primary and multiple

DP START Oslo
2004
FK demultiple

Easy
Cheap

 Requires regular offsets

 Poor on near offsets - amplitude effects
 Suffers from aliasing

DP START Oslo
2004
 Radon Transform
• Radon transform is a way of transforming data from the x-t domain into the Tau-p
domain
• Radon transform is a multi-channel process which involves summing amplitudes
along events in the x-t domain to transform them into the Tau-p domain
• The trajectories along which the amplitudes are summed can be Linear,
Parabolic or Hyperbolic. Depending on the type of transformation, we give the
names:

• Linear Radon Transform

• Parabolic Radon Transform (PRT)
• Hyperbolic Radon Transform
(Each converts data to ‘Tau-p domain’)
DP START Oslo
2004
 Tau-p Domain
X P = -200 -100 0 100 200 300

+p

-200
-100
0
T
+100

+200
+300

 = t - px

DP START Oslo
2004
Tau-p Domain
P=0
A’
A +p

B’

C
C’
Hyperbolae in T-X map to
ellipses in  - p
B
Linear events in T-X map to
single points in  - p

Plane wave
 = t - px

DP START Oslo
2004
Tau-p Domain
• So what is Tau and what is P

• Tau represents intercept time at zero offset

• P represents dip. The larger the dip of an event in the x-t domain the
higher its P value in the Tau-P domain. Flat events (our primaries) appear
around P=0 in the Tau-P domain.
•P = t /  x.

DP START Oslo
2004
 Conversion examples - Diagram
•The zero offset time of the slant path is Tau
•The slant (moveout, 1/velocity, slowness) is P
•A slant whose time increases with offset is +P
•A slant whose time decreases with offset is -P
P
+p 0 -p
Traces (Offset) p1 p2 p3

tau1 Slant = p2 tau1

(tau1, p2)
Slant = p1
tau2 tau2
Time

Tau (tau2, p3)

Slant = p3
(tau3, p1)
tau3 Slant = p2 tau3
(tau3, p2)
DP START Oslo
2004
Tau-p Domain

Several processes are achievable in the tau-p domain, the

most common being -

 Velocity filtering - events with different dips in T-X will

have different ‘p’ values which can be muted/filtered.
 Demultiple –Parabolic Tau-p transform is more commonly
used for multiple attenuations.
 Deconvolution – more effective due to short-period
reverberations being more periodic (for any one p value)

DP START Oslo
2004
PRT Demultiple

Multiple attenuations via PRT is generally accomplished in the following way:

• Input CMP gathered data with NMO applied (primary events
corrected with multiples under-corrected)
• Transform the data into Tau-p domain
• Sum along parabolas as opposed to straight lines
• Primary and multiple energy more focused than in F-K
domain
• Multiples muted prior to transform back to t-x domain or,
primaries are muted and the multiples transformed back to
time domain for subtraction from original data

DP START Oslo
2004
 PRT Demultiple

x x 0 p

NMO PRT

t t Mute


DP START Oslo
2004
PRT Parameters

Let’s look a bit more closely at the parameters for prt:

• MOVEOUT RANGE: Minimum and Maximum times define a
range for conversion to tau-p domain specified at the
reference (maximum) offset
• P TRACES:This range is split into evenly spaced parabolas
which is defined by the number of P-traces specified in the
setup
• MUTE:Area in the tau-p domain that is going to be muted ie
area with either primaries or multiples

DP START Oslo
2004
Moveout Range - writeup

DP START Oslo
2004
Number of P traces

The default number of p-traces is calculated as follows:

Np = 2 (Dtmax - Dt min) f max
where
– Dtmax = Maximum moveout (seconds)

– Dtmin = Minimum moveout (seconds)

– fmax = Maximum frequency (Hz)

– With reference offset set to max offset for gather

DP START Oslo
2004
Moveout Range and P Traces
Zero Reference
offset offset
• This example shows 7 p-
traces
• The 7 parabolas all start at
the same time for zero offset
• They finish at equi-distant
times which span the
Moveout Range

DP START Oslo
2004
Mute Range
Zero Reference
offset offset
• In this case we are muting
the primaries (red area)
and leaving the multiples
to be subtracted from the
original data

DP START Oslo
2004
 theory2.ppt

Data Example
NMO corrected data Multiple Model Primaries

DP START Oslo
Multiples model is subtracted from the original gather
2004
 theory2.ppt

Stacked Data Example

Before demultiple After demultiple

DP START Oslo
2004
PRT Demultiple

Easy
Reasonably AVO friendly

 Expensive
 Can be difficult to parameterize
 Suffers from aliasing and inversion artifacts

DP START Oslo
2004
x-t Predictive Deconvolution

• Apply predictive deconvolution in time to remove

periodic energy

Cheap

 Poor on far offsets where period is not constant

 Statistical

DP START Oslo
2004
-p Predictive Deconvolution

• Apply predictive deconvolution in  on each p trace

to remove periodic energy

Period is constant for horizontal layering

 Cost
 Period changes with p
 Statistical

DP START Oslo
2004
Wave-Equation Multiple Attenuation (WEMA)

• Involves the prediction of water-bottom multiples by wave-

field extrapolation - implemented in the FK domain
• Multiple model subtracted from the input data
• Does not predict other types of multiple e.g. inter-bed
• Requires specification of water depths and water velocity

DP START Oslo
2004
 Wave Equation Multiple Attenuation (WEMA)

Propagate

DP START Oslo & subtract

2004
 Wave Equation Multiple Attenuation (WEMA)

 Only requires knowledge of water layer

 AVO preserving

 Requires simple water bottom

 Only removes water column multiples

DP START Oslo
2004
 At the end of this Session, you will be able to:

• Define the terms primary energy and multiple energy as used in data
processing.
• Define the terms multiple order and period as used in processing.
• Define the following water bottom multiple, free surface multiple, peg-leg
multiple and internal multiple
• Produce a diagram that shows the travel path of each of the following
multiple types: water bottom multiple, free surface multiple, peg-leg
multiple, internal multiple

DP START Oslo
2004
 At the end of this Session, you will be able to:

• Describe how to distinguish primaries from multiple events using their

velocity.
• Describe the removal of multiples using stack.
• Describe the removal of multiples using fk demultiple.
• Describe the transforming of events into the tau-p domain.
• Describe the removal of mutliples using tau-p demultiple.
• Define the following terms PRT, p traces, moveout range and mute range.
• Describe the removal of multiples using wave equation demultiple.

DP START Oslo
2004