Review of Coherent Noise Suppression Methods Gerard T

Review of Coherent Noise Suppression Methods Gerard T. Schuster University of Utah

Problem: Ground Roll Degrades Signal 0 2000 Offset (ft) 3500 Time (sec) Reflections 2. 5 Ground Roll

Problem: PS Waves Degrade Signal Time (sec) 0 4. 0 Reflections Converted S Waves

Time (sec) Problem: Tubes Waves Obscure PP 4. 0 0 2000 Depth (ft) Reflections Time (s) Reflections Aliased tube. Converted waves S Waves 0. 14 3100

Problem: Out-of-Plane Ground Roll

Outline • • Coherent Filtering Methods ARCO Field Data Results Multicomponent Data Example Conclusion and Discussion

Traditional Filtering Methods F-K Dip Filtering in - p domain linear - p parabolic - p hyperbolic - p Least Squares Migration Filter

Separation Principle: Exploit Differences in Moveout & Part. Velocity Directions Time SIGNAL Frequency SIGNAL Transform NOISE Distance NOISE Overlap Signal & Noise Wavenumber

Tau-P Transform Tau Time Sum Transform Distance P

Transform Distance Tau Time Tau-P -P Transform Tau P

Transform Distance Tau Time Tau-P -P Transform Tau Mute Noise P

Transform Tau Time Tau-P Transform Problem: Indistinct Separation Signal/Noise Distance P

Transform Distance Tau Time Hyperbolic Transform Tau-P Transform Distinct Separation Signal/Noise P

v v v v v Irregular Moveout B Time * Breakdown of Hyperbolic Assumption A Distance

Time B A Distance Time Filtering by Parabolic - p Signal/Noise Overlap p

Filtering by LSMF Invert for m p & m s Kirchhoff P-reflectivity Modeler Time PP PS Distance d = Lp mp + L m s s

Filtering by LSMF -1 Time PP Ls Z -1 PS Distance Lp M 21 X

LSMF Method 1. d=L m + L m p p s s data 2. unknowns Find mp by conjugate gradient 3. Model Coherent Signal d = Lp mp

Multicomponent Filtering by LSMF PS PP Time PP Z PS Distance d x = L pmp + L m s s d z = L pmp + L m

Summary Traditional coherent filtering based on approximate moveout LSMF filtering operators based on actual physics separating signal & noise Better physics --> Better focusing, more $$$

Outline • • Coherent Filtering Methods ARCO Surface Wave Data Multicomponent Data Example Conclusion and Discussion

ARCO Field Data Time (sec) 0 2. 5 2000 Offset (ft) 3500

LSM Filtered ARCO Data Field(V. Data Const. ) Time (sec) 0 2. 5 2000 Offset (ft) 3500

F-K Filtered LSM Filtered Data (13333 ft/s) (V. Const. ) Time (sec) 0 2. 5 2000 Offset (ft) 3500

S. F-X of Frequency (Hz) 0 50 Spectrum of ARCO Data LSM Filtered Data (V. Const) F-K Filtered Data (13333 ft/s) 2000 Offset (ft) 3500

Outline • • • Coherent Filtering Methods ARCO Field Data Results Multicomponent Data Example Graben Example Mahogony Example • Conclusion and Discussion

Graben Velocity Model 0 Depth (m) 0 X (m) V 1=2000 m/s V 2=2700 m/s V 3=3800 m/s V 4=4000 m/s V 5=4500 m/s 3000 5000

Synthetic Data 0 Time (s) 5000 0 0 Offset (m) PP 1 PP 2 PP 3 PP 4 5000 PP 3 PP 4 1. 4 Horizontal Component Vertical Component

LSMF Separation 5000 0 Offset (m) 5000 Offset (m) PP 1 Time (s) PP 2 PP 3 PP 4 1. 4 Horizontal Component Vertical Component

True P-P and P-SV Reflection Offset (m) 5000 0 Time (s) 5000 0 0 Offset (m) 1. 4 Horizontal Component Vertical Component

F-K Filtering Separation 0 Time (s) 0 0 5000 Offset (m) PP 1 PP 2 PP 3 PP 4 5000 Offset (m) PP 3 PP 4 1. 4 Horizontal Component Vertical Component

Outline • • • Coherent Filtering Methods ARCO Field Data Results Multicomponent Data Example Graben Example Mahogony Field Data • Conclusion and Discussion

CRG 1 Raw Data Time (s) 0 PS PS 4 CRG 1 (Vertical component) PS

CRG 1 Data after Using F-K Filtering Time (s) 0 PS PS 4 CRG 1 (Vertical component) PS

CRG 1 Data after Using LSMF Time (s) 0 PS PS 4 CRG 1 (Vertical component) PS

CRG 2 Raw Data (vertical component) Time (s) 0 4 CRG 2 (Vertical component)

CRG 2 Data after Using F-K Filtering (vertical component) Time (s) 0 4 CRG 2 (Vertical component)

CRG 2 Data after Using LSMF (vertical component) Time (s) 0 4 CRG 2 (Vertical component)

Outline • Coherent Filtering Methods • ARCO Field Data Results • Multicomponent Data Example • Conclusion and Discussion

Conclusions Filtering signal/noise using: moveout difference & particle velocity direction - Traditional filtering $ vs $$$$ LSMF computes moveout and particle velocity direction based on true physics.