SEISMIC INVERSION
BASICS, PROCEDURE, &
LIMITATION
Seismic section and inversion
BASICS
The estimation of subsurface physical properties by observed
geophysical data is known as “geophysical inversion”,
Inversion technique can be used to any geophysical data like
Gravity, magnetic and seismic etc.
Usually we use seismic data for the inversion
Seismic Inversion create a model of earth's physical property by
using the seismic data as input.
Seismic inversion can be applied “AFTER SATCK” and “BEFORE
STACK”
Both inversions have totally different assumptions. Inversion After
stack based on Zero offset assumption and Before stack multi-
offset assumption.
BASICS
ACOUSTIC IMPEDANCE INVERSION
Normally known as seismic inversion
Applied on processed stacked data i.e. on seismic section
AMPLITUDE VS OFFSET
Normally known as AVO
Applied on pre stack data i.e. on CMP gathers
A new inversion technique Hybrid of AI inversion and AVO is also
working, known as Elastic Inversion
FORWARD MODELLING
Inversion is inverse of forward modeling
For Inversion understanding of Forward modeling is important
Practical example of Forward Modeling is generation of synthetic
seismogram from sonic and density log.
FORWARD MODELLING
SYNTHETIC
DENSITY IMPEDENCE SEISMIC
LOG LOG RC SERIES SEISMIC
SONIC LOG WAVELET
R1
v1 d1 A1
vn dn An Rn
VELOCITY (v) DENSITY (d) (A2 - A1)
A=v xd RC = WL S= RC * WL
(A2 + A1)
INVERSION
Inversion is carried out in following steps
This step is similar to
deconvolution.
Input Data (Seismic Trace)
First we extract or calculate
seismic wavelet.
Conversion of Seismic
STEP-1 trace into Reflectivity
Then its inverse.
Series (RC) By deconvolving this inverse
with trace we get RC
Conversion of reflectivity
STEP-2 series into Acoustic
Impedance log
Conversion of AI log into
STEP-2 Velocity log (Sonic)
Conversion of Sonic into
other attributes like,
Porosity etc.
INVERSION
SEISMIC SEISMIC INVERSE IMPEDENCE DENSITY PSEUDO
TRACE WAVELET WAVELET RC SERIES LOG LOG SONIC LOG
R1 A1 d1 v1
Rn An dn vn
S= RC * WL WL WL-1 RC = S * WL-1 A2 1+RC1 DENSITY (d) VELOCITY (v)
=
A1 1-RC1
TYPE OF AI INVERSION
ACOUSTIC IMPEDANCE
INVERSION
Deterministic Inversion Probabilistic or Stochastic
Inversion
MODEL SPARSE RECURSIVE
BASED SPIKE INVERSION
INVERSION INVERSION
Sparse Spike and Model Based inversions are commonly used in
industry
Sparse spike inversion is also known as Constrained Sparse Spike
Inversion
SPARSE SPIKE
INVERSION
SPARSE SPIKE INVERSION
ASUMPTIONS
Based on assumption that major lithological boundaries represent
large events in reflectivity series.
These large events are superimposed on smaller events (noise)
Larger events are the spikes on reflectivity series
Theoretical procedure of SSI is given below
We start with initial estimate of wavelet 0 WAVELET
ESTIMATE
Calculate Reflectivity
Improve the wavelet
Calculate Reflectivity
This procedure is done in iterative way ESTIMATE
IMPROVE
SPARSE
WAVELET
REFLECTIVITY
For Detail Ref: Chapter-6, Introduction to seismic
inversion methods, Brain H. Russel
SPARSE SPIKE INVERSION
PARACTICAL WORKING
Estimation of Wavelet, its Inverse and Reflectivity Series (RC)
Inversion of RC
Generation of low frequency model
Addition of low frequency model into inversion results
Conversion to Inversion results into porosity
SPARSE SPIKE INVERSION
ESTIMATION OF WAVELET Wavelet Estimation
Iterative process Schematic
Generate RC from well (s) data
Extract wavelet from seismic
Generate Synthetic trace
Seismic line
Compare Synthetic with seismic through well
Calculate error
Wavelet extraction
Update wave let
Repeat process until we get
minimum error
WELL DATA
SONIC & Trace from
DENSITY LOGS Well
ERROR
RC Series Synthetic Trace
Comparison
SPARSE SPIKE INVERSION
Trace RC
ESTIMATION OF RC SERIES
We have a Wavelet
Calculate its Inverse
Convolve this inverse with all
seismic data
Commonly known as stratigraphic
de-convolution
Now all traces converted to RC
Series
Inverse
Wavelet
=
*
SPARSE SPIKE INVERSION
INVERSION AND POROSITY CALCULATION
From RC series we calculate Acoustic Impedance (AI) logs
A2 1+RC1
=
A1 1-RC1
AI is converted to velocity logs (AI = V x D)
Now all data traces are in velocity logs i.e. Pseudo Sonic logs
From these sonic logs we can calculate porosity.
SPARSE SPIKE INVERSION
ADDITION OF LOW FREQUENCY MODEL
Why we add low frequency model
Low frequencies are missing from seismic data as
We do not record them due to ground roll problem
Eliminated during processing due to noise
Limitation of the recording geophones
Absence of low frequency can show false impedance layers on the
final data
Inversion without low frequency
Full Band Inversion
SPARSE SPIKE INVERSION
ADDITION OF LOW FREQUENCY MODEL
Generation of low frequency model
Low frequencies calculated from
Filtered Sonic log (s)
Seismic velocity Analysis
From geological model
Each method has its limitation, but normally a combination of sonic
and seismic velocity data used
Low frequency can be added either at reflectivity or impedance
stage
Adding at reflectivity stage is good
MODEL BASED
INVERSION
MODEL BASED INVERSION
PARACTICAL WORKING
Estimation of Wavelet
Generation of initial Impedance model
Updating Impedance model after comparing synthetic and original
seismic
Final model is one which have least error.
MODEL BASED INVERSION
Impedance model building
It is based on picked time horizons and well log data
First we define horizontal Impedance layers changing w.r.t. wells
Then these layer divided into small cells making grid
Every grid node has a value of time and Impedance
The value of Impedance can be constant or linear changing vertically
Commonly a vertically changing layer is used
After every step this grid model updated and node values changed
Impedance layers based on wells Initial grid of impedance layers
Time horizon from seismic data and time horizons
MODEL BASED INVERSION
SSI vs. MBI
SPRASE SPIKE INVERSION MODEL BASED INVERSION
Seismic data itself used Seismic data not directly used
Chances of getting non Chances of getting non
geologically output are low geologically output are high
Noise is severe problem Noise is not as severs as in SSI
Resolution is low Resolution is high
Multiples can use as false layers Multiples are easily handled
Low frequency model have to be Low frequency model is included
added separately in the algorithm
SSI MBI
LIMITATION OF INVERSION
There are two major limitation of inversion method
Non uniqueness
Thin bed tuning
Non uniqueness mean one can get many answer for one seismic data
Seismic data
If we have
Different impedance models to fit same
X + Y = 10 seismic
Different values of X & Y can fit to get 10
4 + 6 = 10
5 + 5 = 10
7 + 3 = 10
So many
LIMITATION OF INVERSION
How to handle non uniqueness
Defining the constraints i.e. geological input or common
sense about the model
Seismic data
If we have
We can say our model has thick sand unit
X + Y = 10 with some %age of shale
We can say X & Y are even, greater than 2
and less than 8
4 + 6 = 10
then we have we have only one solution
LIMITATION OF INVERSION
Thin bed Tuning
Tuning is the interference effect of the
seismic reflection from top and bottom
of the thin bed
It causes artificial changes in amplitude
hence Impedance
Inversion results
Seismic data showing showing artificial
True Impedance model tuning effect impedance
