Reflection Traveltime Tomography Using Dynamic Smoothing

Download Report

Transcript Reflection Traveltime Tomography Using Dynamic Smoothing

First Arrival Traveltime and
Waveform Inversion of
Refraction Data
Jianming Sheng and Gerard T. Schuster
University of Utah
October, 2002
Outline
• Motivation
• First arrival traveltime and
waveform inversion
• Numerical examples
• Summary
Motivation
Given: Traveltime and waveform
of CDP refraction data
Goal: High resolution tomogram
Problem: Can waveform tomography
provide better resolution than
ray-based tomography?
Ray-based Tomography vs.
Full Waveform Inversion
Ray-based
tomography
Efficient and robust
Full
waveform
tomography
No high-freq. limitation
Resolution limited by
high-freq. assumption
Slow convergence and
local minima problem
First-arrival Traveltime and
Waveform Inversion
Ray-based
traveltime
tomography
Efficient and robust
First-arrival
waveform
inversion
No high-freq. limitation
Initial model
Better convergence and
mild nonlinear
Outline
• Motivation
• First arrival traveltime and
waveform inversion
• Numerical examples
• Summary
First Arrival Traveltime and
Waveform Inversion
• Step 1: Preprocessing the raw data:
band-pass, 3D to 2D
transform, trace normalization
• Step 2: Picking first-arrival
traveltimes and muting out
other waves except first
arrivals
• Step 3:
First arrival traveltime
tomography
Minimizes traveltime residual



obs
Initial model


pre 2
• Step 4: First arrival
waveform inversion
Predicted
Observed
Misfit
function

s,g
obs
s,g
D
D
pre
s,g 2
Multigrid Tomography
• Traveltime tomography:
Dynamic smoothing scheme
(to attack local minima problem)
(Nemeth, T., Normark, E. and Qin, F., 1992)
Outline
• Motivation
• First arrival traveltime and
waveform inversion
• Numerical examples
• Summary
Numerical Examples
• Synthetic data I: Three-layer
• Synthetic data II: WesternGeco
(Blind test)
• Redmond mine survey data
Synthetic Model I
0
2500
20
1958
Suggested by
Source
Freq.
60
Hz
Konstantin Osypov
40
1416
Avg. Velocity 2400 m/s
873
Source Wavelength 40 m
60
0
100
Distance (m)
200
331
(m/s)
Synthetic Model I
0
2500
20
1958
1416
40 m
40
873
60
0
100
Distance (m)
200
331
(m/s)
Synthetic Data I
•
•
•
Synthetic data set was calculated
by 2-D FD acoustic wave equation
solver
Twenty-one shots and 51 traces
per shot were used.
Computational grid dimension was
401*121.
Synthetic Shot Gather
-80
Time (sec.)
0.0
Offset (m)
0.1
Air Wave
120
Traveltime Tomogram
0
2500
20
1958
1416
40
873
60
0
100
Distance (m)
200
331
(m/s)
Synthetic Model I
0
2500
20
1958
1416
40
873
60
0
100
Distance (m)
200
331
(m/s)
Traveltime Residual
2.0
1.0
0.0
1
Iterations
30
Waveform Tomogram
0
2500
20
1958
1416
40
873
60
0
100
Distance (m)
200
331
(m/s)
Synthetic Model I
0
2500
20
1958
1416
40
873
60
0
100
Distance (m)
200
331
(m/s)
Waveform Residual
12,000
8,000
4,000
0
1
Iterations
30
Numerical Examples
• Synthetic data I: Three-layer
• Synthetic data II: WesternGeco
(Blind test)
• Redmond mine survey data
True Velocity Model
0.0
Depth (km)
0.0
1.0
Horizontal distance (km)
26
1000 m/s
2050~2500 m/s
True Density Model
0.0
Depth (km)
0.0
1.0
Horizontal distance (km)
26
Recorded CSG # 150
Time (sec.)
-3000
0.0
2.0
Offset (m)
3000
Guessed Density Model
Density (kg/m3)
3400
 v 
  2100 

 1981.2 
0.5
1400
1000
Velocity (m/s)
5000
Source Wavelet
Amplitude
400
0
-600
0.0
Time (sec.)
0.25
Waveform Matching
Offset
(m)
-50
Amplitude
-25
0
25
50
0.0
Time (sec.)
0.2
Traveltime Tomogram
0.0
0.0
Horizontal distance (km)
26
m/s
2712
Depth (km)
2284
1856
1428
1.0
1000
Traveltime Tomogram
Depth (km)
5.0
0.0
Horizontal distance (km)
8.75
m/s
2409
0.1
2057
0.2
1705
0.3
1352
0.4
1000
Waveform Tomogram
Depth (km)
5.0
0.0
Horizontal distance (km)
8.75
m/s
2700
0.1
2275
0.2
1850
0.3
1425
0.4
1000
Migration section
Depth (km)
5.0
0.0
0.1
0.2
0.3
0.4
Horizontal distance (km)
8.75
Predicted CSG #150
Time (sec.)
-3000
0.0
2.0
Offset (m)
3000
Recorded CSG # 150
Time (sec.)
-3000
0.0
2.0
Offset (m)
3000
Numerical Examples
• Synthetic data I: Three-layer
• Synthetic data II: WesternGeco
(Blind test)
• Redmond mine survey data
Salt Diapir Data
•
Thirty-one shots and 120 traces
total 3188 traveltimes picked.
Shot interval: 20 m
geophone interval 5 m
•
Source frequency 40 Hz.
•
Record length 1 sec.
sample interval 0.5 millisecond .
CSG for Field Data
After Preprocessing
Time (sec.)
0
0.2
1
Geophone #
120
CSG for Field Data
After Muting
Time (sec.)
0
0.2
1
Geophone #
120
Wavelet Extracted
0
0.1
Traveltime Tomogram
5500
0
20 m
4500
3500
55 m
SALT
130
2500
Tunnel
1500
500
0
Distance (m)
590
(m/s)
Traveltime Residual
2.0
1.0
0.0
1
Iterations
30
Waveform Tomogram
5500
0
20 m
4500
3500
55 m
SALT
130
2500
Tunnel
1500
500
0
Distance (m)
590
(m/s)
Traveltime Tomogram
5500
0
20 m
4500
3500
55 m
SALT
130
2500
Tunnel
1500
500
0
Distance (m)
590
(m/s)
Waveform Residual
6,000
4,000
2,000
0
1
Iterations
30
Predicted CSG
Time (sec.)
0
0.2
1
Geophone #
120
CSG for Salt Data
After Muting
Time (sec.)
0
0.2
1
Geophone #
120
Logarithmic Amplitude Vs. Offset
Log10 Amplitude
2
0
Synthetic
-2
Observed
-4
0
Offset (m)
400
Problems
Seismic attenuation
Surface wave noise
Source wavelet inversion
& objective function
Outline
• Motivation
• First arrival traveltime and
waveform inversion
• Numerical examples
• Summary
Summary
• Synthetic results show that the
waveform tomogram is much more
resolved;
• The preliminary results for the field
data are not as good as expected, and
further work is needed.
Acknowledgment
I thank the sponsors of the 2002 University
of Utah Tomography and Modeling
/Migration (UTAM) Consortium for their
financial support . I thank Konstantin
Osypov for providing the data set.