Wave-Equation Migration Velocity Analysis Paul Sava and Biondo Biondi * Stanford Exploration Project Stanford University EAGE 2004 Workshop on Velocity biondo@stanford.edu
Deep-water subsalt imaging 2 1) Potentials of wavefield-continuation methods can be fulfilled only if we use MVA methods based on: Wavefield-continuation migration • Salt-boundary picking • Below salt Common Image Gathers (CIG) Wavefield-continuation velocity updating 2) We may need to go beyond downward-continuation migration methods and … be able to perform MVA biondo@stanford.edu
Deep-water subsalt imaging - Velocity problem? 3 1) Potentials of wavefield-continuation methods can be fulfilled only if we use MVA methods based on: Wavefield-continuation migration • Salt-boundary picking • Below salt Common Image Gathers (CIG) Wavefield-continuation velocity updating 2) We may need to go beyond downward-continuation migration methods and … be able to perform MVA biondo@stanford.edu
Deep-water subsalt imaging - I ll umination? 4 1) Potentials of wavefield-continuation methods can be fulfilled only if we use MVA methods based on: Wavefield-continuation migration • Salt-boundary picking • Below salt Common Image Gathers (CIG) Wavefield-continuation velocity updating 2) We may need to go beyond downward-continuation migration methods and … be able to perform MVA biondo@stanford.edu
“Simple” wavepath with f=1 26 Hz 5 biondo@stanford.edu
“Complex” wavepath with f=1 26 Hz 6 biondo@stanford.edu
“Messy” wavepath with f=1 26 Hz 7 biondo@stanford.edu
“Messy” wavepath with f=1 3 Hz 8 biondo@stanford.edu
“Messy” wavepath with f=1 5 Hz 9 biondo@stanford.edu
“Messy” wavepath with f=1 12 Hz 10 biondo@stanford.edu
“Messy” wavepath with f=1 16 Hz 11 biondo@stanford.edu
“Messy” wavepath with f=1 26 Hz 12 biondo@stanford.edu
Wavepaths in 3-D 13 biondo@stanford.edu
Wavepaths in 3-D – Banana or doughnuts? 14 biondo@stanford.edu
Velocity Analysis and wavefield methods 15 Brief history of velocity estimation with wavefield methods • Full waveform inversion (Tarantola, 1984, Pratt, today) • Diffraction tomography (Devaney and Oristaglio, 1984) • Wave-equation tomography (Woodward, 1990; Luo and Schuster 1991) • Differential Semblance Optimization (Symes and Carazzone, 1991) Challenges of velocity estimation with wavefield methods • Limitations of the first-order Born linearization (“Born limitations”) • Problems with large (in extent and value) velocity errors • Dependent on accurate amplitudes both in the data and in the modeling • Computational and storage requirements of explicit use of wavepaths biondo@stanford.edu
Velocity information in ADCIGs - Correct velocity 16 Sources Receivers γ 1 γ 1 γ 2 γ 3 γ 2 γ 3 α V mig = V true Depth biondo@stanford.edu
Velocity information in ADCIGs - Low velocity 17 Sources Receivers Δ l 2 Δ l 1 Δ l 3 Δ l 1 < Δ l 2 < Δ l 3 V mig < V true γ 1 γ 1 γ 2 γ 3 γ 2 γ 3 α V mig < V true Depth biondo@stanford.edu
Ray-tomography Migration Velocity Analysis 18 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ z 3) Invert Δ z into Δ s by solving: ( ) min W Ä z − L ray s Δ Ä s 2 where L ray is given by raytracing biondo@stanford.edu
Wave-Equation Migration Velocity Analysis 19 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: ( ) min W Ä I − L wave s Δ Ä s 2 where L wave is given by first-order Born linearization of wavefield continuation biondo@stanford.edu
Wave-Equation Migration Velocity Analysis 20 Sava and Biondi 1) (2004) Measure errors in ADCIGs Important! by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: ( ) min W Ä I − L wave s Δ Ä s 2 where L wave is given by first-order Born linearization of wavefield continuation biondo@stanford.edu
Ray tomography MVA Wave-Equation MVA 21 L ray L wave biondo@stanford.edu
Ray tomography MVA Wave-Equation MVA 22 L ray L wave biondo@stanford.edu
Ray tomography MVA Wave-Equation MVA 23 Δ z Δ I L ray L wave biondo@stanford.edu
Deep-water subsalt data 24 1) Potentials of wavefield-continuation methods can be fulfilled only if we use MVA methods based on: Wavefield-continuation migration • Salt-boundary picking • Below salt Common Image Gathers (CIG) Wavefield-continuation velocity updating 2) We may need to go beyond downward-continuation migration methods and … be able to perform MVA biondo@stanford.edu
Deep-water subsalt data - Initial velocity 25 biondo@stanford.edu
Deep-water subsalt data - Initial velocity 26 biondo@stanford.edu
Deep-water subsalt data – WEMVA step 1) 27 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: ( ) min W Ä I − L wave s Δ Ä s 2 biondo@stanford.edu
Deep-water subsalt data – WEMVA step 1) 28 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: Δρ = ρ− 1 ( ) min W Ä I − L wave s Δ Ä s 2 biondo@stanford.edu
Deep-water subsalt data – WEMVA step 2) 29 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: Δ I ( ) min W Ä I − L wave s Δ Ä s 2 biondo@stanford.edu
Deep-water subsalt data – WEMVA step 2) 30 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: Δρ = ρ− 1 ( ) min W Ä I − L wave s Δ Ä s 2 biondo@stanford.edu
Deep-water subsalt data – WEMVA step 3) 31 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: Δ I ( ) min W W Ä I − L wave s Δ Ä s 2 W biondo@stanford.edu
Deep-water subsalt data – WEMVA step 3) 32 1) Measure errors in ADCIGs by measuring curvature ( ρ ) 2) Convert measured ρ into Δ I 3) Invert Δ I into Δ s by solving: s 0 + Δ s ( ) min W Ä I − L wave s Δ Ä s 2 s 0 biondo@stanford.edu
Deep-water subsalt data – Initial velocity 33 biondo@stanford.edu
Deep-water subsalt data – Velocity after 2 iterat. 34 biondo@stanford.edu
Deep-water subsalt data – Initial image 35 Image biondo@stanford.edu
Deep-water subsalt data – Image after 2 iterat. 36 Image biondo@stanford.edu
Deep-water subsalt data – Initial ADCIGs 37 ADCIGs biondo@stanford.edu
Deep-water subsalt data – ADCIGs after 2 iterat. 38 ADCIGs biondo@stanford.edu
Deep-water subsalt data – Initial ADCIGs 39 ADCIGs biondo@stanford.edu
Deep-water subsalt data – ADCIGs after 2 iterat. 40 ADCIGs biondo@stanford.edu
Deep-water subsalt data – Initial Δρ Δρ = ρ -1 41 Δρ = ρ -1 Δρ White flat ADCIGs biondo@stanford.edu
Deep-water subsalt data – Δρ Δρ after 2 iterations 42 Δρ = ρ -1 Δρ White flat ADCIGs biondo@stanford.edu
Deep-water subsalt data – W after 2 iterations 43 Weights White reliable ρ picks biondo@stanford.edu
Conclusions 44 • Ray-based Migration Velocity Analysis (MVA) methods have been successful in complex structure, but they are challenged by subsalt velocity estimation. biondo@stanford.edu
Conclusions 45 • Ray-based Migration Velocity Analysis (MVA) methods have been successful in complex structure, but they are challenged by subsalt velocity estimation. • Wave-equation MVA (WEMVA) can be accomplished while preserving the work-flow of conventional ray-based MVA methods. biondo@stanford.edu
Conclusions 46 • Ray-based Migration Velocity Analysis (MVA) methods have been successful in complex structure, but they are challenged by subsalt velocity estimation. • Wave-equation MVA (WEMVA) can be accomplished while preserving the work-flow of conventional ray-based MVA methods. • The velocity function estimated by the use of our WEMVA method results in flatter ADCIGS and more coherent reflectors, even if we started from a high-quality velocity function that was estimated with ray-based MVA. biondo@stanford.edu
Conclusions 47 • Ray-based Migration Velocity Analysis (MVA) methods have been successful in complex structure, but they are challenged by subsalt velocity estimation. • Wave-equation MVA (WEMVA) can be accomplished while preserving the work-flow of conventional ray-based MVA methods. • The velocity function estimated by the use of our WEMVA method results in flatter ADCIGS and more coherent reflectors, even if we started from a high-quality velocity function that was estimated with ray-based MVA. • Poor illumination prevents the extraction of reliable velocity information from ADCIGs at every location, and thus presents a challenge also for WEMVA. biondo@stanford.edu
Recommend
More recommend
