Mathematics of Kirchhoff Inversion One-day Short Course presented by Norman Bleistein University Emeritus Professor Department of Geophysics Center for Wave Phenomena Colorado School of Mines norm@dix.mines.edu cwp.mines.edu/~norrm Presented Sunday, September 11, 2005 International Meeting Sociedada Brasileira de Geofísica Salvador, Brazil Preface In 2001, Mathematics of Multidimensional Seismic Imaging, Migration and Inversion (MMSIMI), by Jack K. Cohen, John W. Stockwell and me, was published by Springer-Verlag. Since that time, I have continued doing research in this general area, largely as a consequence of a consulting relationship with Veritas, DGC. Not surprisingly, my viewpoint about the underlying theory has evolved since the publication of that book. In particular, I prefer starting the inversion derivation from an asymptotic form of the forward model of the reflected wave as a volume integral with a Reflectivity function “nearly” completely identified in that representation. Furthermore, the two-and-one-half-dimensional theory would seem to have been overtaken by the speed and memory capacity of computers, so that I have now relegated it to a last part of the current presentation. Taking higher priority in my mind is an extension of the Kirchhoff theory that starts from an integral inversion formula with the variables of integration being coordinates at the image point. The formula starts relatively compactly stated, although implementation presents problems. However, when the integral is recast in source/receiver coordinates, all sorts of salutary consequences arise. Among them: i. Division by Green’s function amplitudes in traditional Kirchhoff inversion formulas now is replaced by multiplications by adjoint Green’s functions. This provides for more stable numerical integration formulas. ii. The inversion is a sum over all sources and receivers that provides an output in a suite of common angle panels, covering both opening angles and azimuthal angles between specular rays to sources and receivers from the image point. iii. The sum over available source/receiver pairs—available traces—lends itself to relatively straightforward processing of data collected from surveys using multi-streamer receiver sets. The classical Kirchhoff inversion method does not provide a formula to process data acquired in this manner. The seminal ideas for this new work must be credited to Sheng Xu and Yu Zhang of Veritas, DGC, with Samuel H. Gray (Veritas) and Guanquan Zhang providing challenges, inspiration and astute criticism and oversight as these ideas developed. (An incomplete bibliography is listed below.) Preparing these slides was a useful exercise for me. First, it forced me to acknowledge my changed priority order and point of view. Second, it forced me to attempt a uniformity of notation that certainly has not been my hallmark in the past. Some of the figures addressing ideas of the text have been revised and there are new figures developed along with the new ideas mentioned above. I just hope there is enough time in the day! Norm Bleistein July 28, 2005 Bleistein, N., Zhang, Y., Xu, S., Gray, S. H., Zhang, G., 2005, Migration/Inversion: Think Image Point Coordinates, Process in Acquisition Surface Coordinates: preprint, http://cwp.mines.edu/~norm/Papers/MIUpperSurf.pdf Xu, S., Chauris, H, Lambaré G. and Noble, M. S., 2001, Common-angle migration: A strategy for imaging complex media: Geophysics, 66, 6, 1877-1894. Zhang, Y., Xu, S., Zhang, G., Bleistein, N., 2004, How to obtain true amplitude common-angle gathers from one-way wave equation migration?: Expanded Abstracts, International Meeting of the Society of Exploration Geophysicists , Tulsa.. i. Short Course, Brazil Outline 1. Motivation—forward modeling and imaging o Hagedoorn modeling o Hagedoorn imaging Analytical description o Mathematical description of a reflector Singular function of a surface o Wave equation Time domain Frequency domain Green’s functions o Kirchhoff modeling, single reflector, exact o Plan for deriving asymptotic inversion o Effect of band limiting on imaging 2. Ray theory, propagation, reflection o “Ansatz” Eikonal equation (travel time) Transport equation (Amplitude) Initial data on rays • Green’s function • Reflection 3. Asymptotic modeling of reflected wave o Method of stationary phase o Kirchhoff approximations applied to exact Kirchhoff formula o Recast surface integral formula as a volume integral 4. Deriving Kirchhoff inversion o Formal inversion via approximate (local) Fourier transform Beylkin determinant Properties of local Fourier interpretation Special cases • Common shot • Common offset 5. Full wave form Kirchhoff inversion o Motivation: Better Green’s function ⇒ better imaging o Review of “true amplitude” concept o Approximate pseudo-inverse applied to Kirchhoff exact modeling formula Kirchhoff approximation applied to better Green’s functions Modeling formula recast as a volume integral • Interpretation as a modeling operator o Inversion Adjoint of modeling operator Pseudo-inverse of modeling operator Asymptotic normalization of operator Relation of result to Kirchhoff inversion with ray-theoretic Green’s functions 6. Think image point coordinates, process in source/receiver coordinates o Motivation o Inversion as an integral over angular variables at the image point o Transformation to acquisition surface coordinates Deconvolution processing ⇒ Correlation processing o Inversion of data acquired from multi-streamer surveys 7. Two-and-one-half-dimensional (2.5D) inversion o Motivation (3D processing over a 2D (cylindrical) Earth model) o 3D inversion for identical lines of data Stationary phase applied to out-of-plane integration in operator In-plane inversion • In-plane 3D propagation • Application to in-plane inversion o 2.5D inversion formula Kirchhoff (Born) Inversion Short course Norm Bleistein Center for Wave Phenomena Colorado School of Mines cwp.mines.edu/mmsimi 2 1 More current point of view than in MMSIMI Migration/Inversion: Think Image Point Coordinates, Process in Acquisition Surface Coordinates: with Y. Zhang, S. Xu, S. H. Gray, G. Zhang, to appear • cwp.mines.edu/~norm Kirchhoff inversion is… • Leading order • High frequency • Asymptotic technique 4 2 So is … • Wave equation migration –(including “true amplitude”) • Kirchhoff migration • Stolt (f/k) migration • Phase shift migration 5 Graphic approach to inversion …due to Hagedoorn 3 7 8 4 Record arrival times • Source/receiver position - ξ • Arrival time - t ? • Arrival time - vt/2 10 5 11 12 6 13 14 7 15 16 8 17 18 9 19 Hagedoorn construction 20 10 Kirchhoff processing 21 22 11 23 D Data at time τ on trace ξ 24 12 25 D Data at time τ on trace ξ W Spatial weighting function “Reflectivity” 26 13 D Data at time τ on trace ξ D Filtered observed data • Wave shaping 27 W Spatial weighting function • Migration (Reflector map) • Inversion (Parameter estimation) • Model consistent 28 14 “Reflectivity” To be defined later 29 Course objective: Fill in the details 30 15 Reflector image Singular function of a surface Delta function of distance normal to the surface Singular function of a surface: 16 Reflector image: image of the singular function of the reflector Action of the singular function Volume integral Surface integral 17 Tools of the trade Forward modeling Forward modeling The wave equation 18 Forward modeling The wave equation The wave equation Frequency domain 38 19 Green’s function 39 Point source 40 20 Green’s function representation of solution 41 Volume part 42 21 Surface part 43 Born modeling • Volume integral • Pseudo-sources – Created from medium perturbations 22 Kirchhoff modeling • Surface integral – Reflector • Ray theory/ WKBJ for data – Generalization of the Kirchhoff approximation Kirchhoff modeling 23 Inversion objective • Image the reflector • Estimate reflection coefficient – Plane waves, planar reflectors – Curved reflectors? • “High frequency concept!” • Length scales >> wave length “High frequency” modeling 24 “High frequency” inversion “High frequency” inversion with offset 25 “The Plan” • Model reflected wave field – data acquisition • Common shot, common offset, etc • Identify “Reflectivity” – 51 “The Plan” • M reflected wave field – data acquisition • Common shot, common offset, etc • Identify “Reflectivity” – • “Solve” for Reflectivity 52 26 Why reflectors? Not even “steps” in the in velocity, density High frequency data cannot resolve steps! 54 27 High frequency data cannot resolve steps! 55 High frequency data cannot resolve steps! 56 28 High frequency data cannot resolve steps! 57 High frequency data cannot resolve steps! 58 29 High frequency data can resolve delta functions! 59 Kirchhoff modeling We don’t know 30 Kirchhoff modeling We don’t see in Kirchhoff modeling We need ray theory for incident waves, reflected waves 31 Summary • Hagedoorn construction – Integral inversion formula • Forward modeling as point of departure • Reflector imaging –High frequency inversion 63 32 Tools of the Trade Ray theory/ WKBJ • “High frequency” concept – plane waves, planar reflectors – length scales large compared to a wave length 64 Homogeneous wave equation Ansatz 65 1 Homogeneous wave equation First term in an asymptotic series 66 Substitute …. 67 2 Eikonal equation 68 First order nonlinear partial differential equation 69 3 Remember …. 70 (First) Transport equation 71 4 Tools of the Trade Ray theory/ WKBJ • “Ansatz” • Eikonal equation (phase, traveltime) • Transport equation (amplitude) 72 Solution technique: Method of characteristics 73 5 Solution technique: • Solve for trajectories in space, rays • Determine τ along the rays 74 75 6 76 Alternative ray parameter σ Important in 2.5D 77 7 Ray parameter, s, arclength 78 An example: — point source — constant wave speed 79 8 Example: reflected wave 80 Example: reflected wave • Surface: 81 9 Example: reflected wave • Surface: • Traveltime: 82 Example: reflected wave • Surface: • Traveltime: • Gradient: solve 83 10 84 Tools of the Trade Ray theory/ WKBJ • “Ansatz” • Eikonal equation (phase, traveltime) • Transport equation (amplitude) 85 11 (First) Transport equation 86 Integrate over a “smart” volume 87 12 88 Preserved on the ray Preserved on the ray Preserved on the ray 89 13 Preserved on the ray Preserved on the ray Preserved on the ray 90 Preserved on the ray Preserved on the ray 91 14 92 Amplitude propagation 93 15 Other parameters along the ray 94 Green’s function amplitude C: depends on choice of α1, α2 95 16 Example: α1, α2, polar coordinates 96 Reflection Formulation 97 17 Wave types • Incident wave • Reflected wave • Transmitted wave 98 Principle: Wave incident on the reflector gives rise to reflected and transmitted waves. 99 18 Total wave field • Above reflector • Below reflector 100 Objective Find initial data on S for reflected and transmitted wave in terms of incident wave Express τR, τT, AR, AT, in terms of τI and AI 101 19 Interface conditions on S Wave field and its normal derivatve are continuous 102 103 20 104 Determines tangential components of gradient 105 21 Normal components? Eikonal equation 106 107 22 Amplitudes 108 Amplitudes 109 23 Important for modeling and inversion! 110 What happens when J = 0? Caustic! 111 24 What happens when J = 0? Caustic! 112 Ray theory fails at caustic • Higher function required – Airy function • Phase shift through caustic 113 25 More caustics? • KMAH index, κ – Count of the number of caustics 114 Summary: ray theory • Wave equation • Describes solution with amplitude and phase: Aeiωτ • τ: Eikonal equation • A: Transport equation • Green’s function • Reflection process 115 26 Asymptotic modeling of reflected wave • Method of stationary phase • Kirchhoff approximations (from ray theory) applied to exact Kirchhoff formula • Recast surface integral formula as a volume integral 116 Further asymptotics Method of stationary phase 117 1 Prototype integral Formally: 118 Prototype integral Practically: λ “large” 119 2 Prototype integral Forward modeling 120 Prototype integral Migration/Inversion 121 3 Prototype integral Migration/Inversion 122 Prototype integral Stationary points 123 4 Contribution from stationary point 124 Contribution from stationary point Count: pos e-values - neg e-values 125 5 How accurate? 126 One dimension: n = 1 127 6 Example: the Hankel Function Green’s function in 2D, homogeneous medium (within a constant) 128 Example: the Hankel Function 129 7 Real part: solid - exact 130 Imaginary part 131 8 Percentage error 132 Return to modeling equation 133 9 Use ray theory to simplify 134 135 10 136 137 11 138 139 12 Generalization of The Kirchhoff Approximation 140 141 13 Use ray theory to simplify 142 Use ray theory to simplify 143 14 Write as volume integral? Use singular function γ(x)? Not with normal derivatives 144 Write as volume integral? Analyze: method of stationary phase 145 15 Stationary phase analysis 146 Stationary phase analysis 147 16 Stationary phase analysis 148 Stationary phase analysis 149 17 Stationary phase analysis Snell’s law! 150 151 18 Stationary phase analysis Snell’s law! 152 Stationary phase analysis Snell’s law! 153 19 Stationary phase analysis Snell’s law! 154 Revised reflection modeling 155 20 Revised reflection modeling stationary phase approximation 156 Revised reflection modeling Surface integral to volume integral 157 21 Revised reflection modeling Reflectivity function! 158 “The Plan” • Model reflected wave field – data acquisition • Common shot, common offset, etc • Identify “Reflectivity” – 159 22 Revised reflection modeling Finally!!!! 160 Summary: arriving at • Start from Kirchhoff integral • Use ray theory to simplify • Use stationary phase to simplify • Transform surface integral to volume integral 161 23 Deriving Kirchhoff inversion 162 Revised reflection modeling 163 1 Objective: invert for reflectivity 164 Sources and Receivers Common shot Common offset 165 2 Phase 166 Amplitude 167 3 Same modeling operator, Different notation Inverse? 168 Cascade modeling and inversion operators 169 4 `must be a delta function! 170 `must be a delta function! 171 5 `must be a delta function! 172 ‘must be a delta function! 173 6 What about the Jacobian? 174 Beylkin determinant 175 7 Beylkin determinant 176 Solve! 177 8 Recall 178 Kirchhoff inversion h accomodates variable upper surface 179 9 Equivalence • Start from Born approximation for forward modeling • Same inversion formula 180 Micro-local geometry at the image point 181 10 Micro-local: changes with source point 182 Micro-local: changes with receiver point 183 11 Micro-local: changes with image point 184 Micro-local geometry at the image point 185 12 Micro-local geometry at the image point 186 Beylkin determinant: alternative form 187 13 Beylkin determinant: alternative form 188 Beylkin determinant: alternative form 189 14 Common Shot (planar acquisiion surface) xs = constant, xr = (ξ1, ξ2,0) 190 Common Shot (planar acquisiion surface) xs = constant, xr = (ξ1, ξ2,0) 191 15 Common Offset • No simplification • Computationally intensive 192 Common Offset • Geometric computation 193 16 Summary • Inversion formula for reflectivity –Unspecified source/receiver specification –Two acquisition surface parameters –Characterized by Beylkin determinant 194 What can we prove? • Apply inversion formula to (Kirchhoff) model data • Two integrals for modeling, three for inversion • Apply multi-dimensional stationary phase (spatial coordinates) 195 17 x Stationary triple for each x 196 197 18 198 199 19 200 201 20 202 θ stat = θ spec 203 21 x = xstat 204 Stationary value of gradient Specular value of gradient 205 22 Second inversion operator 206 207 23 Other attributes: same trick • Specular source, receiver • Specular traveltimes • Specular point • Surface normal at image point 208 Summarize • Two reflectivities • Peak values • Peak ratio 209 24 Accuracy 2D, 2.5D • Peak values: 2% • Peak ratio: .2% 210 Accuracy 3D • Peak values: 4% • Peak ratio: ?% 211 25 Kirchhoff inversion • Start from forward model – Volume integral, single reflector • Locally, Fourier transform – Formally invert • Reasonable accuracy on tests 212 26 Kirchhoff inversion with full wave form Green’s functions To build a better image… Use a better Green’s function! (Better than ray theory.) 213 Forward model • Kirchhoff approximation (Better Green’s function) 214 1 Inversion • Kirchhoff approximation (Better Green’s function) 215 Reflection coefficient Plane waves Planar reflectors Homogeneous media 2 Reflection coefficient, High frequency asymptotics Curved wave fronts (but not too … .) Curve reflectors (but not too … .) Heterogeneous media (but not too … .) (but not too … .) • f Frequency in hz • v “Average” wave speed • L Length scale 218 3 (but not too … .) L Length scale: – Radius of curvature of wave front – Radius of curvature of reflector – 219 Return to exact forward model 220 4 Return to exact forward model Kirchhoff approximation 221 Kirchhoff approximation 222 5 Full wave form Kirchhoffapproximate modeling 223 ..with some more asymptotics 224 6 …as a volume integral 225 Singular function of a surface: 7 …as a volume integral 227 …as a volume integral 228 8 …as a volume integral • Gaussian beams • “True amplitude” one-way • Full wave equation 229 …as a volume integral 230 9 …as a volume integral 231 …as a volume integral 232 10 Solve via pseudo-inverse adjoint to operator 233 Operates on functions of and produces functions of Operates on functions of and produces functions of 234 11 How does it work? 235 Operates on functions of and produces functions of 236 12 Note: no bandlimiting in Replace 237 Recall 238 13 “True amplitue” principles • Asymptotic concept • Requires single arrivals • Only need asymptotically • Ray theory, stationary phase 239 Asymptotic Earlier inversion 240 14 Inversion for reflectivity 241 Compare to earlier inversion Same “filter” Same weight New! 242 15 Inversion for reflectivity 243 Note! New result becomes old result 244 16 Summary • Start from exact Kirchhoff formula with exact or “better quality asymptotic Green’s functions • Apply asymptotics to obtain volume integral representation • Derive a formal pseudo-inverse for reflectivity 245 17 Think Image Point Coordinates Process in Source/Receiver Coordinates Norman Bleistein Yu Zhang Sheng Xu Guanquan Zhang Samuel H. Gray 246 Why? • Image point inversion for simpler Beylkin determinant • Acquisition surface for easier trace acquisition • More stable processing 247 1 Why image point coordinates? 248 Kirchhoff Inversion deconvolution structure (including image point Kirchhoff M/I) Image Point Kirchhoff M/I, upper surface processing correlation structure 249 2 Kirchhoff Inversion deconvolution structure (including image point inversion) Image Point Kirchhoff M/I, upper surface processing correlation structure 250 Kirchhoff Inversion deconvolution structure (including image point inversion Image Point Kirchhoff inversion, upper surface processing correlation structure 251 3 ⇒ Reflection data Fourier domain information about medium 252 k = ω grad traveltime 253 4 k = ω∇τ ( x, xs , xr ) 254 Micro-local: changes with source point 255 5 Micro-local: changes with receiver point 256 Micro-local: changes with image point 257 6 2cosθ k =ω ν̂ v( x) 258 2cosθ k =ω ν̂ v( x) 259 7 Integrate in Fourier Variables { 3 ⎡ 2cosθ ⎤ ∂ν̂ ∂ν̂ dk1dk2 dk3 = ω ⎢ ⎥ × dω dν1dν 2 v( x) ∂ν ∂ν ⎣ ⎦ 1 2 2 260 Integrate in Fourier Variables { 3 ⎡ 2cosθ ⎤ ∂ν̂ ∂ν̂ dk1dk2 dk3 = ω 2 ⎢ ⎥ × dω dν1dν 2 ⎣ v( x) ⎦ ∂ν1 ∂ν 2 Beylkin determinant 261 8 Integrate in Surface Variables (Midpoints, Receivers) 3 ⎡ 2cosθ ⎤ ∂ν̂ ∂ν̂ dk1dk2 dk3 = ω ⎢ ⎥ × dω dξ1dξ2 v( x) ∂ξ ∂ξ ⎣ ⎦ 1 2 2 262 Integrate in Surface Variables (Midpoints, Receivers) 3 ⎡ 2cosθ ⎤ ∂ν̂ ∂ν̂ dk1dk2 dk3 = ω 2 ⎢ ⎥ × dω dξ1dξ2 ⎣ v( x) ⎦ ∂ξ1 ∂ξ2 Beylkin determinant 263 9 Regular surface patch, irregular sphere patch 264 Regular sphere patch, irregular surface patch 265 10 3 3 ⎡ 2cosθ ⎤ ∂ν̂ ∂ν̂ ⎡ 2cosθ ⎤ × = ⎥ sin ν1 ⎢ v( x) ⎥ ∂ν ∂ν ⎢ v( x) ⎣ ⎦ ⎣ ⎦ 1 2 266 Kirchhoff Inversion in Image Point Coordinates 1 2 cosθ R( x,θ , φ ) = 4π 2 v( x) D( x, xr , xs ) ∫ A( x, xs )A( x, xr ) sin ν1dν1dν 2 267 11 Kirchhoff Inversion in Image Point Coordinates 1 2 cosθ R( x,θ , φ ) = 4π 2 v( x) D( x, xr , xs ) ∫ A( x, xs )A( x, xr ) sin ν1dν1dν 2 A’s: WKBJ Green’s function amplitudes 268 Kirchhoff Inversion in Image Point Coordinates 1 2 cosθ R( x,θ , φ ) = 4π 2 v( x) D( x, xr , xs ) ∫ A( x, xs )A( x, xr ) sin ν1dν1dν 2 1 iω u( xr , xs , ω ) D( x, xr , xs ) = ∫ 2π • exp{−iωτ ( x, xr , xs ) + iκ sgn(ω )π / 2}dω 269 12 1 iω u( xr , xs , ω ) D( x, xr , xs ) = ∫ 2π • exp{−iωτ ( x, xr , xs ) + iκ sgn(ω )π / 2}dω • u( xs , xr , ω ) : Data 270 1 i ω u( x D( x, xr , xs ) = r , xs , ω ) 2π ∫ • exp{−iωτ ( x, xr , xs ) + iκ sgn(ω )π / 2}dω • u( xs , xr , ω ) : Data • κ ( x, xs , xr ) : KMAH index 271 13 1 iω u( xr , xs , ω ) D( x, xr , xs ) = ∫ 2π • exp{−iωτ ( x, xr , xs ) + iκ sgn(ω )π / 2}dω • u( xs , xr , ω ) : Data • κ ( x, xs , xr ) : KMAH index • τ ( x, xs , xr ) : Travel time 272 Kirchhoff Inversion in Image Point Coordinates 1 2 cosθ R( x,θ , φ ) = 4π 2 v( x) D( x, xr , xs ) ∫ A( x, xs )A( x, xr ) sin ν1dν1dν 2 Deconvolution processing 273 14 Kirchhoff Inversion in Image Point Coordinates 1 2 cosθ R( x,θ , φ ) = 4π 2 v( x) D( x, xr , xs ) ∫ A( x, xs )A( x, xr ) sin ν1dν1dν 2 Sorts in common angle gathers 274 Kirchhoff Inversion in Image Point Coordinates 1 2 cosθ R( x,θ , φ ) = 4π 2 v( x) D( x, xr , xs ) ∫ A( x, xs )A( x, xr ) sin ν1dν1dν 2 xs , xr determined by ray tracing 275 15 Change variables to upper surface coordinates 276 ν ,ν ,two variables, 1 2 xs , xr , four variables 277 16 Trick: θ = ∫ δ (θ '− θ )dθ ' and φ = ∫ δ (φ '− φ )dφ ' ‘ ‘ ‘ 278 Kirchhoff Inversion in Image Point Coordinates ∫ R( x, θ ,φ ) = 1 4π 2 2 cosθ ' D( x, xr , xs ) ∫ v( x) A( x, xs )A( x, xr ) • δ (θ '− θ )δ (φ '− φ )sin ν1dν1dν 2 dθ 'dφ ' 279 17 Change variables of integration ν1 , ν 2 , θ ', φ ' ⇒ α s1 , α s 2 , α r1 , α r 2 ‘ ‘ ‘ 280 Change variables of integration α s1 , α s 2 ⇒ xs1 , xs 2 α r1 , α r 2 ⇒ xr1 , xr 2 ‘ ‘ ‘ 281 18 Change variables of integration α s1 , α s 2 ⇒ xs1 , xs 2 α r1 , α r 2 ⇒ xr1 , xr 2 Ray Jacobians! 282 Change variables of integration α s1 , α s 2 ⇒ xs1 , xs 2 α r1 , α r 2 ⇒ xr1 , xr 2 Ray Jacobians! Green’s function amplitudes! 283 19 Kirchhoff Inversion in Source/Receiver Coordinates cos β cos β R( x,θ , φ ) = 32π 2 v( x) ∫ s r D( x, xs , xr ) v( xs ) v( xr ) • A* ( x, xs )A* ( x, xr )δ (θ '− θ )δ ( sin θ (φ '− φ )) • dxs1dxs 2 dxr1dxr 2 284 Kirchhoff Inversion in Source/Receiver Coordinates cos β cos β R( x,θ , φ ) = 32π 2 v( x) ∫ s r D( x, xs , xr ) v( xs ) v( xr ) • A* ( x, xs )A* ( x, xr )δ (θ '− θ )δ ( sin θ (φ '− φ )) • dxs1dxs 2 dxr1dxr 2 Correlation processing! 285 20 Kirchhoff Inversion in Source/Receiver Coordinates cos β cos β R( x,θ , φ ) = 32π 2 v( x) ∫ s r D( x, xs , xr ) v( xs ) v( xr ) • A* ( x, xs )A* ( x, xr )δ (θ '− θ )δ ( sin θ (φ '− φ )) • dxs1dxs 2 dxr1dxr 2 Multi-streamer data acquisition 286 Full Wave Form Kirchhoff Migration/Inversion cos β cos β R( x,θ , φ ) = 16π v( x) ∫ s r iω u R ( xs , xr ,ω )dω v( xs ) v( xr ) • G * ( xs , x,ω )G * ( xr , x,ω )δ (θ '− θ ) • δ ( sin θ (φ '− φ )) dxs1dxs 2 dxr1dxr 2 287 21 Full wave form - true amplitude • Reflection coefficient : plane wave, WKBJ concept 288 Full wave form - true amplitude • Reflection coefficient : plane wave, WKBJ concept • G* : ~ WKBJ Green’s function 289 22 Full wave form - true amplitude • Reflection coefficient : plane wave, WKBJ concept • G* : ~ WKBJ Green’s function • Full wave form Kirchhoff M/I: true amplitude processing 290 Image Point Kirchhoff M/I • Single-valued map to sources/receivers 291 23 Image Point Kirchhoff M/I • Single-valued map to sources/receivers • Simple transformation to surface coordinates 292 Image Point Kirchhoff M/I • Single-valued map to sources/receivers • Simple transformation to surface coordinates • Deconvolution processing ⇒ Correlation processing 293 24 Image Point Kirchhoff M/I • Single-valued map to sources/receivers • Simple transformation to surface coordinates • Deconvolution processing ⇒ Correlation processing • Relatively simple formulas for computation 294 Image Point Kirchhoff M/I • Output in common angle gathers 295 25 Image Point Kirchhoff M/I • Output in common angle gathers • Ideal for multi-streamer survey data 296 Image Point Kirchhoff M/I • Output in common angle gathers • Ideal for multi-streamer survey data • Full wave form Kirchhoff inversion 297 26 Two-and-one-half Dimensions(?!) 2.5D A concept whose importance has been overtaken by computer capacity 298 Two-and-one-half Dimensions(?!) Single line processing 3D Green’s functions over a “2D” earth (Hence, 2.5D) 299 1 2.5D acquisition 300 Two-and-one-half Dimensions(?!) • Single line of data • Best acquired in the dip direction 301 2 Two-and-one-half Dimensions(?!) • Physics: “nearly” no variation in out-of-(x,z)-plane direction • Mathematics: no variation in out-of -(x,z)-plane direction – Velocity independent of y (or x2) 302 303 3 Two-and-one-half Dimensions(?!) Operator depends on out-of- plane variable Data does not depend on out-of- plane variable Precompute out-of-plane integral asymptotically 304 Recall Frequency domain integral isolated 305 4 Equivalent 306 Data independent of ξ2 307 5 Data independent of ξ2 308 Velocity independent of y (or x2) 309 6 Apply stationary phase in ξ2 Output independent of x2 310 Apply stationary phase in ξ2 Propagation is in the vertical (x1,x3) plane containing x2 311 7 Only need in-plane values of 312 Need: in-plane 3D ray theory 313 8 Recall: ray equations with parameter σ 314 Ray equations with parameter σ 315 9 Ray equations with parameter σ 316 Ray equations with parameter σ 317 10 Ray equations with parameter σ 318 Ray equations with parameter σ • Rays are in-plane 2D propagation, 3D geometrical spreading 319 11 Ray equations now 2D with parameter σ 320 Recall: amplitude, 3D 321 12 2.5D in-plane amplitude 322 2.5D in-plane amplitude out-of-plane spreading in-plane spreading 323 13 2.5D in-plane amplitude 324 2.5D in-plane amplitude 325 14 Beylkin determinant 326 Beylkin determinant, 2.5D 327 15 Apply stationary phase in ξ2 328 2.5D inversion formula 329 16 The data, D2D 330 2.5D Summary • Single line of data in dip direction • Assume v = v(x,z) • 3D inversion with identical lines of data (independent of y) • Apply stationary phase in out-of-plane direction • 2.5D inversion formula 331 17