SEISMIC DATA ANALYSIS Processing, Inversion, and Interpretation of Seismic Data OZ YILMAZ Volume I Stephen M. Doherty, Editor Society of Exploration Geophysicists Post Office Box 702740, Tulsa, OK 74170-274012 13 a 15 TABLE OF CONTENTS PREFACE TO THE FIRST EDITION PREFACE Volume 1 INTRODUCTION Processing of Seismie Data, 4 Inversion of Seismic Data, 10 Interpretation of Seismic Data, 18 From Seisinie Exploration to Seismic Monitoring, 22 Chapter 1 FUNDAMENTALS OF SIGNAL PROCESSING Introduction, 25 ‘The 1-D Fourier Transform, 26 Analog versus Digital Signal, 28 Frequency Aliasing, 30 Phase Considerations, 34 Time-Domain Operations, 36 Convolution, 38 Crosscorrelation and Autocorrelation, 39 Vibroscis Correlation, 41 Frequency Filtering, 41 Practical Aspects of Frequeney Pitering, 4 Bandwidth and Vertical Resolution, 46 Time-Variant Filtering, 48 ‘The 2D Fourier Transform, 48 Spatial Aliasing, 51 Worldwide Assortment of Shot Records, 67 Wave Types, 70 Gain Applications, 81 Geometric Spreading Correction, 81 Programmed Gain Control, 85 RMS Amplitude AGC, 85 Instantancous AGC, 87 Relative Trace Balancing, 89 Basic Data Processing Sequence, 90 Preprocessing, 91 Deconvolution, 92 COMP Sorting, 93 Velocity Analysis, 93 Normal-Moveout Correction, 94 Multiple Attenuation, 94viii Seismic Data Analysis Dip-Moveout Correction, 94 OMP Stacking, 95 Poststack Processing, 95 Migration, 95 Residual Staties Corrections, 122 Quality Control in Processing, 122 Parsimony in Processing, 124 Exercises, 150 Appendix A: A Mathematical Review of the Fourier Transform, 153 A.1 The L-D Fourier Transform, 153 A2 The ¢-Transform, 155 A3 The 2-D Fourier Transform, 156 References, 156 Chapter 2 DECONVOLUTION 2.0 Introduction, 159 2.1 The Convolutional Model, 162 ‘The Convolutional Model in the Time Domain, 167 ‘The Convolutional Model in the Frequency Domain, 170 2.2 Inverse Filtering, 171 ‘The Inverse of the Source Wavelet, 172 Least-Squares Inverse Filtering, 173 Minimum Phase, 175 2.8 Optimum Wiener Filters, 179 Spiking Deconvolution, 180 Prewhitening, 181 Wavelet Processing by Shaping Filters, 183 Predictive Deconvolution, 185 2.4 Predictive Deconvolution in Practice, 190 Operator Length, 190 Prediction Lng, 193 Percent Prewhitening, 203, Effect of Random Noise on Deconvolution, 207 Multiple Attemaation, 209 2.5 Field Data Examples, 211 Prestack Deconvolution, 213 Signature Deconvolution, 217 Vibroseis Deconvolution, 219 Poststack Deconvolution, 2 2.6 ‘The Problem of Nonstationarity, 222 ‘Time-Variant Deconvolution, 227 ‘Time-Variant Spectral Whitening, 231 Frequency-Domain Deconvolution, 233 Inverse @ Filtering, 234 Deconvolution Strategies, 241 Exercises, 247 ‘Appendix B: Mathematical Foundation of Deconvolution, 249 B.1 Synthetic Seismogram, 249 B.2 The Inverse of the Source Wavelet, 251 B.3 The Inverse Filter, 252 BA Frequency-Domain DeconvolContents ix BS Optimum Wiener Filters, 186 Spiking Deconvolution, 258 187 Predictive Deconvolution, 260 1B Surfer: Consistent Deconvolution, 262 19 Inverse Q Filtering, 266, References, 270 Chapter 3 VELOCITY ANALYSIS AND STATICS CORRECTIONS, 8.0 Introduction, 271 BL Normal Moveout, 274 NMO for a Fla Reflector, 274 [NMO in » Horizontally Stratified Barth, 280, Fourth-Order Movwont, 280 NMO Stactching, 283 NMO for a Dipping Reflector, 285, [NMO for Several Layers with Arbitrary Dips, 287 oveout Velocity versus Stacking Velocity, 288 8.2 Velocity Analysis, 288, ‘The Velocity Spectrum, 202 Measite of Cohereney, 295 Factors Affecting Velocity Bstimates, 902 Interactive Velocy Analysis, 311 Horizon Velocity Analysis 312 Cohereney Attribute Stacks, 318, 3.3 Residual Staties Corrections, 324 Resin! States Fstination by Traveltime Decomposition, 396 ‘Residual Statics Estimation by Stack-Power Maximization, 344 “Teaveltime Decomposition in Practice, 45, ‘Maximum Allowable Shift, 346 Corteation Window, 361 (Other Considerations, 362 Stack Power Masinization in Praetic, 305 SA Refraction Statics Corrections, 370 First Breaks, 374 Fed States Corrections, $75, Flat Refractor, 375 Dipping Refractor, 377 ‘The Phis Mins Method, 377 ‘The Generalized Reciprocal Method, $79, ‘Th Least Squares Method, 379 ‘Processing Sequence for Statics Corrections, $81 ‘Model Expertnents, 382 Field Data Examples, 295 Bxereises, 432 Appendix C: Topies in Moveout and Statics Corrections, 437 ‘C1 The Shifted Hiyperbola, 437 (C2 Moveout Stretch, 420 €.3 Equations for a Dipping Reflector, 441 C44 Traveltime Decomposition for Residual Statics Pstimation, 442 C5 Depth Entimation from Refracted Arvvals, 444 66 Equations for a Dipping Refractor, 445Seismic Data Analysis .7 The Plus-Minus Times, 447 C.8 Generalized Linear Inversion of Refracted Arrivals, 448 C.9 Refraction Traveltime Tomography, 453 C.10 L,-Norm Refraction Statics, 456 References, 460 4.0 4.2 43 44 45 Chapter 4 MIGRATION Introduction, 463, Exploding Reflectors, 467 Migration Strategies, 470 Migration Algorithms, ATL Migration Parameters, 474 Aspects of Input data, 475 tation Velocities, 475 Migration Principles, 476 Kirchhoff Migration, 481 Diffraction Summation, 484 Amplitude and Phase Factors, 485, Kirchhoff Summation, 485 Finite-Difference Migration, 486 Downward Continnation, 486 Differencing Schemes, 488 Rational Approximations for Implicit Schemes, 489 Reverse Time Migration, 491 Frequency-Space Implicit Schemes, 492 Proquency-Space Explicit Schemes, 493 Frequency-Wavenumber Migration, 494 Phase-Shift Migration, 498 Stolt Migration, 500 Summary of Domains of Migration Algorithms, 501 Kirchhoff Migration in Practice, 502 Aperture Width, 502 Maximum Dip to Migrate, 509 Velocity Errors, 509 Finite-Difference Migration in Practice, 520 Depth Step Size, 521 Velocity Errors, 525 Caseaded Migration, 525 Reverse Time Migration, 930 Frequency-Space Migration in Practice, 530 Steep-Dip Implicit Methods, 535, Depth Step Size, 537 Velocity Errors, 344 Steep-Dip Explicit Methods, 549 its of Extrapolation Fi Errors, 552 Frequency-Wavenumber Migration in Practice, 559 Maximum Dip to Migrate, 559 Depth Step Size, 566 Velocity Errors, 587 Stolt Stretch Factor, 572 ers, 549Contents ‘Wraparound, 575 Residual Migvation, 3% 4.6 Further Aspects of Migration in Practice, 579 ‘Migration and Spatial Aiasing, 581 Migration and Random Noise, 619 Migration and Line Length, 621 Migration from Topography, 626 Exorcisos, 626 “Appendix D: Mathematical Foundation of Migration, 628 ‘Dal Wavefield Extrapolation and Migration, 628 D2 Stationary Phase Approximations, 638 D4 The Parabolic Approximation, 699 Dat Froqueney-Space Implicit Schemes, 611 D.5 Stable Explicit Extrapolation, 614 ‘D6 Optimum Depth Step, 646 D7 Froqueney-Wavenutnber Migration, 619 D& Residual Migration, 651 References, 652 Chapter 5 DIP-MOVEOUT CORRECTION AND PRESTACK MIGRATION 5.0 Introduction, 655 SleFlank Reflections, 657 Fault Plane Reflections, 657 DMO and Stacking Velocities, 657 ‘Taming Wave Reflections, 665, 5.1 Principles of Dip-Moveout Correction, 668 Prestack Paria} Migzation, 670 Frequeacy-Wavenumber DMO Correction, 672 Log-Stretch DMO Correction, 677 Integral DMO Correction, 679 Velocity Errors, 681 Vasiahle Velociy, 684 Tuming- Wave Migration, 685 5.2 Dip-Moveout Correction in Practice, 692 Salt Flank, 692 Fale Planes, 693 DMO and Mutples, 705, DMO and Coherent Linear Noise, 716 Other Considerations, 716 Aspects of DMO Corteetion — A Summary, 722 5.8 Prestack Time Migration, 725 DMO Comrection and Coumon-Offset Migration, 728, Salt Plaks, 720 Fault Planes, 742 Comanot-Reflection-Point versus Common-Refletion Surface Stacking, 769 ‘Migration Velocity Analysis, 775 Prestaek Stalt Migration, 70 ‘Common-Offet Migration of DMO-Cosrected Dato, 7 restock Kirchbot Migration, 788 Velocity Analysis Using Common-Relletion-Point Gathers, 788 Focusing Analysis, 798 ‘Fowler's Velocity-independent Prestack Migration, 803 xixii Soismic Data Analysis Bxorcises, 815 Appendix Es Topics in Dip-Moveout Correction and Prestack Time Migration, 817. E.1 Reflection Point Dispersal, 817 E.2 Equations for DMO Correction, 820 E.3 Log Stretch DMO Correction, 823 E.l The DMO Ellipse, 828 E.5 Nonzero-Offset Traveltime Equation, 827 E.G Prostack Frequeney-Wavenmmber Migration, 831 ET Velocity Analysis by Wavefield Extrapolation, 833 References, 834 Chapter 6 NOISE AND MULTIPLE ATTENUATION 6.0 Introduction, 837 Coherent Linear Noise, 838 ‘Treatment of Coberent Linear Noise by Conventional Processing, 840 Roverberations aud Multiples, 843 ‘Treatment of Reverberations and Multiples by Conventional Process Spatially Random Noise, 870 6.1 Multiple Attenuation In the CMP Domain, 877. Periodicity of Multiples, 877 Velocity Discrimination Between Primaries and Multiples, 887 Karlnueu-Loove Transform, 887 Moxleling of Multiples, 896 6.2 Frequency-Wavenumber Filtering, 898 Random Noise and Frequency-Wavenumber Filtering, 94 Statics Corrections and Frequeney-Wavenuniber Filtering, 905 Dip Filtering of Coherent Linear Noise, 905 Frequency-Wavenumber Multiple Attenuation, 907 6.8 The Slant-Stack Transform, 920 Physical Aspects of Slant Stacking, 920 Slant-Stack Transformation, 923, Practical Aspects of Slant Stacking, 924 Slant-Stack Parameters, 028 ‘Time-Variant Dip Filtering, 981 Slant-Stack Multiple Atvennation, 982 6.4 The Radon Transform, 938 Volocity-Stack Transformation, 942 ‘The Discrete Radon Transform, 943 ‘The Parabolic Radon Transform, 44 Practical Considerations, 915 Impulse Response of the Velocity-Stack Operator, 948 Field Data Examples, 048 Radon-Transform Multiple Attemation, 953 6.5 Linear Uncorrelated Noise Attenuation, 960 Design of Spatial Prediction Filters, 966, Field Data Examples, 966 Exercises, 976 Appendix F: Multichannel Filtering Techniques for Noise and Multiple Attenuation, 977 F.L Analysis of Guided Waves, 977 F.2 Wavefield Extrapolation in the r ~ p Domain, 980 F.3 Mathematical Foundation of the Discrete Radon Transform, 982 87Contents 4 Free Surface Multiple Attonnation, 989 P55 Water-Bottom Multiple Attenuation, 92 E56 Spatial Priction Filter, 95, References, 998 INDEX, xxv Volume It Chapter 7 &-D SEISMIC EXPLORATION 1.0 Introduction, 1001 "The Need for Imaging in Thee Dimensions, 1008, TA 8-D Survey Design and Acquisition, 1010 Migration Aperture, 1010, Spatial Sampling, 1017 Other Consierations, 1018 ‘Marine Acquisition Geometry, 1018, Cale Fathering, 1019 3D Binning, 1019 ‘Crossline Staring, 1020 Strike versus Dip Shooting, 1027 Land Acquisition Geometr 7.2 Processing of 8-D Seismic Data, 1030 SD Retraction Statics Corrections, 1036 [Azimuth Dependence of Moveout Velocities, 1096 SD Dip-Moveout Corretion, 1016 Inversion to Zero Offset, 1015, Aspeets of 3D DMO Correction — A Summary, 1050 Velocity Analysis, 1050 SD Residual States Corrections, 1050 SD Migration, 1051 ‘Trace Interpolation, 1065 7.8 8-D Poststack Migration, 1073 Separation versus Splitting, 1073 Impulse Response of the One-Pass Implicit Finite Difference 3-D Migration, 1074 ‘Two-Pass versus One-Pass Implicit Fiaite Difference SD Migration in Practice, 1076 Explicit Sehemes Combined with the MeClelian Transform, 1082 ‘The Phase Shift-Phs-Correstion Method, 1088, TA &D Prestack Time Migration, 1099 D DMO Correction Combined with $-D Comuon-Oflse Migeation, 1112 (Crossine Migration, 1129, |:D Migration Velocity Analyst 1131 ‘Aspects of -D Prestack Tinie Sfgration — A Summary, 1137 7.5 Interpretation of S-D Seismic Data, 1158 ‘Time Shces, 1156 3D Visualization, 1156 Removal of Opacity, 1188, Seed Detection, 1150 Strictural Interpretation, 161 Stratigrapie Interpretation, LIT Exoreises, 1195xiv Seismic Data Analysis Appendix G: Mathematical Foundation of 8-D Migration, 1198 G1 Implicit Methods, 1198 G.2 Explicit Methods, 1200 G.3 3-D Phase-Shift Migration, 1208 G.4 3D Stolt Migration, 1204 G.5 Trace Interpolation, 1204 G6 3D Nonzero-Oflset: Traveltime Equation, 1208 References, 1209 Chapter 8 EARTH IMAGING IN DEPTH 8.0 Introduction, 1213 Lateral Velocity Variations, 1222 8.1 Layer Replacement, 1226 Wave-Equation Datuming, 1229 Poststack Layer Replacement, 1230 Prestack Layer Replacement, 1231 Field Data Example, 1237 8.2 2-D Poststack Depth Migration, 1238 Image Rays and Lateral Velocity Variations, 1238 ‘Time versus Depth Migration, 1244 Iterative Depth Migration, 1247 Iteration with Zero-Offsct Data, 1250 Iteration with CMP-Stacked Da Iteration with Prestack Data, 126: Iteration in Practice, 1265 8.3 2-D Prestack Depth Migration, 1273 Shot-Geophone Migration, 1274 Shot-Profile Migration, 1280 usitivity of Image Accuracy to Velocity Errors, 1280, Field Data Examples, 12 8.4 3-D Poststack Depth Migration, 1304 3-D Poststack Time versus Depth Migration, 1304 ‘Two-Pass versiis One-Pass 3-D Poststack Depth Migration, 1313 Implicit versus Explicit 3-D Poststack Depth Migration, 1314 3-D Poststack Datuming, 1321 8.5 3-D Prestack Depth Migration, 1321 Kirchhoff Summation, 1324 Calculation of Traveltimes, 13 ‘The Bikonal Equation, 1325 Fermat's Principle, 1331 Summation Strategies, 1331 Migration Aperture, 1333 Operator Antialiasing, 1333 3-D Common-Offset Depth Migration, 1335 Exercises, 1342 Appendix H: Diffraction and Ray Theory for Wave Propagation, 1342 H.1 The Kirchhoff Integral, 1342 T1.2 The Eikonal Equation, 1346 11.3 Finite-Difference Solution to the Fikonal Equation, 1349 References, 1351Contents xv Chapter 9 BARTH MODELING IN DEPTH 8.0 Introduction, 1853, Inversion Methods for Duta Modeling 1355 Teworson Procedutes for Earth Modeling, 1356 Volocity-Depth Ambiguity, 1857 ‘Model Representation and Visualization, 1360 9.1 Models with Horizontal Layers, 1365 Dix Conversion, 1365, Cohereney Inversion, 1369 [Near Surtnce Layer with Laterally Varying Velocities, 1382 9.2 Model with Low-Relief Structure, 1387 Stacking Velocity Inversion, 1802 (Cohereney Inversion, HL Velocity Resolution, 1404 9.3. Model with Complex Overburden Structure, 1404 Tinnge-Gathers, 1406 Constant Half Space Velocity Analysis, 15 9.4 Model Building, 1415 ‘Time to-Depth Conversion, 1416 ‘Time Stoucture Mapa, L116 Interval Velocity Maps, 1117 Depth Structute Maps, 1425 Calibration to Wel Tops, 1426 Tayer-by-Layer Inversion, LASS Stractre-Ddependent version, 1450, 8.5 Model Updating, 1450 Residual Moveout Analysis, 1462 election Traveltine Toxography, 1469 Limitations in Resolving Velocty-Depth Ambiguity by Tomograpy. 1479, Turning: Ray Tomography, 1512 Exercises, 1524 Appendix J: Data Modeling by Inversion, 1525 {11 The Generalized Linear Inversion, 1525, 4.2 The GLI Formalism of Deconvolution, 1526 13.3 Applications of the GLI Technique, 1530, ‘34 Dix Conversion, 1534 55.5 Map Processing, 1539 {16 Reflection Travetime Tomography, 1545, 437 Threshold for Velosity-Depth Ambiguity, 1553, References, 1554 Chapter 10 STRUCTURAL INVERSION 10.0 Introduction, 1557 10.1. Subsalt Imaging in the North Sea, 1588 [Extimstion ofthe Overburden Model, 1562 [Estimation ofthe Substrarum Model, 1562 Mosel Verifeation, 1963 10.2 Subsalt Imaging in the Gulf of Mexico, 1974 Layered Earth Model Estimation, 1574xvi Seismic Data Analysis Structuse-Independent Model Batimation, 1577 10.3 Imaging Beneath Irregular Water Bottom in the Northwest Shelf of Australia, 1597 Barth Modeling anel Imaging in Depth, 1507 10.4 Imaging Beneath Volcanies in the West of Shetlands of the Atlantic Margin, 1597 Earth Modeling snd Imaging In Depth, 1607 10.5 Imaging Beneath Shallow Gas Anomalies in the Gulf of Thailand, 1620 Earth Modeling and Inieging in Depth, 1620 10.6 8-D Structural Inversion Applied to Seismic Data from the Southern North Sea, 1626 Estimation of the Overburden Model, 1626 Model Representation by Tessellation, 1630 SD Coherency hiversion, 1680 SED Poststack Depth Migration, 1037 Estimation of the Substrattm Model, 1638 10.7 3-D Structural Inversion Applied to Seismic Data from the Central North Sea, 1651 3D Coherency laversion Combined with 3D Poststack Depth Migration, 1665 B.D Stacking Velocity Inversion Combined with 3-D Image-Ray Depth Conversion, 1674 10.8 3-D Structural Inversion Applied to Seismic Data from Offshore Indonesia, 1674 Mode! Building, 1678 Mocie! Updating, 1678 Taging. in Depth, 1690 ‘Volume-Based Interpretation, 1690 10.9 24D Structural Inversion Applied to Seismic Data from the Northeast China, 1703 SD DMO Processing, 1720 B.D Prestack Time Migration, 1720 From RMS to Interval Velocities, 1742 Structural Inversion, 1742 Structural and Stratigraphic Interpretation, 174 Exercises, 1778 Appendix K: Seismic Modeling, 1779 Kl Zero-Offset Traveltime Modeling, 1779 K.2 Zero-Offeet Wavelield Modeling, 1781 K.3 Nonzero-Olfset Wavefield Modeling, 1781 K4 Elastic Wavefield Modeling, 1790, References, 1792 Chapter 11 RESERVOIR GEOPHYSICS 11.0 Introduction, 1793 Blastie Waves and Rock Properties, 1704 11.1 Seismic Resolution, 1801 Vertical Resolution, 1801 Lateral Resolution, 1803 11.2 Analysis of Amplitude Variation with Offset, 1807 Rellection and Refraction, 1808 Reflector Curvature, 1813 AVO Equations, 1816 Processing Sequence for AVO Analysis, 1880 Derivation of AVO Attributes hy Prestack Amplitude Inversion, 1851 Interpretation of AVO Attributes, 1862 3-D AVO Analysis, 1863 11.3 Acoustic Impedance Estimation, 1863 Symthetie Sonie Logs, 1861Contents xvi Processing Sequence for Acoustic npedance Estimation, 1865 Derivation of Acoustic lmpedance Atteibute, 1806 3D Acoustic Impedance Estimation, 1872 Instantanwous Attributes, 1896 11.4 Vertical Seismic Profiling, 1907 VSP Acquisition Geometry. 1907 Processing of VSP Data, 1907 VSP-CDP Trausfori, 1008 11.5 4D Seismic Method, 1911 Processing of 4-D Seismic Data, 1912 Selanic Revervoit Monitoring, 1913 11.6 4-C Seismic Method, 1915 Recording of 1-C Seismic Data, 1919 Gainers Coupling Analysis of Geophone Data, 1922 Processing of PP Data, 1926 Rotation of Horizontal Geophone Components, 1926, CCommon-Conversion-Point Binning, 1033 Velocity Analysis of PS Data, 1946 Dip-Maweont Correction af PS Data, 1959 [Migration of PS Data, 1961 1.7 Seismic Anisotropy, 1961 Anibotropie Velocity Analysis, 1965, Auisotropie Dip-Moveout Correction, 1968 “Anizotronic Migration, 1980 Elect of Anisotropy on AVO, 1998, Sheat-Wave Spiting in Anisotrople Media, 1999, Exercises, 2000 Appendix L: Mathematical Foundation of Elastic Wave Propagation, 2001 TL. Stns-Stmin Relation, 2001 1.2 Blastie Wave Equation, 2007 3 Seismic Wave Types — Body Waves and Surface Waves, 2008 [LA Wave Propagation Phenomena — Diffraction, Reflection, aud Refraction, 2012 L8 The Zocppritz Equations, 2014 Log Prestack Amplitude Laversion, 2019 References, 2024 INDEX, xxvPREFACE TO THE FIRST EDITION “The slanic method playa prominent roe nthe soar, for hydrocarbons, Setanic exploration consists of thee rain stages: data arquisicon, proeslng, nd terpre- tation. This book i intended to hep the seismic analyst understand the fundamentals of the techniques sed i processing seis data In particular, emphasis is given tothe practical aspects of data mays “Toes in hs nol are trated in two phases. Fst, ce process is described fram a physical viewpoint ‘with les emphasis on matheretical development. 1 doing £0, geometric means arc used extensively to help ‘the reader gain the physical insight into the diferent proces, Second, the geophysical parameters that al fect the felity of the esting ontpat from each pro- ccs aro critically examined vin an extensive series of synthetic and real data examples. For the student of = Action sekmology and new entrants to che sels In dustry this book ries to provide insights into the prac teal aspocts ofthe application of the theory of tie so- ves and waves For experienced seismic exploratonists this book should serve as a refesher and andy re sronce. However, it isnot just meant for the sisie fanalyst. Explorsionists who would ike to gn a prac- ‘cal bnekyround in seis data processing without any mathematical burden ak should benefit fom it. Nev frtholes, for the more thooreielly inclined, xmathe- naticl treatise on the main subjects s provided in the appendixes ‘The scismie analyst is confronted daily with the Important tasks of (1) selecting a proper sequence of processing steps ap- propriate for the field data under consideration, (2) Selecting an appropriate set of parameters fe ene processing se ad (3) evaluating the testing output from exch proces ing step then diagnosing any problems eaused by Improper parameter selection. ‘There is a well-established sequence for standard seismic data procissing. The three principal processes ‘deconvolution, stacking, and migration — make wp the foundation of routine processing. There also are same auxiliary proceses tat help improve the eee- tiveness of che priniple process. Quostons often arise sto the kind of ansiney process that shouldbe used tnd when they should be appli For example if sot ecards conan an abundance of source generated co- herent nose, then dip filtering may be valuable before deconvolution. Besin steering may’ be meoasary (0 am Drove the signal-to-noise ratio while reducing the num- ber of channels in procesing bya factor of as much as fou. Residual statieseorrctins often are raul for improving velocity estimation and stacking. In a daly production vironment, many questions arse concern ing the optimal parameter section for each process Somme of the most repeatedly asked questions are: What is guod length forthe deconvolution operator? What should the prediction Ing. be? What should the de- Sign te for the oparntor be? How should the corre. lation window be chocen in residual statis computa tious? What kind of apecture width shoukl one select in Kirchhoff migration? Whats the optimum depth step sleo in nite diffrence migration? Many more questions could be included in ths lst of questions. To help am fer these questions, & large number of examples us Ing both field and synthetic data and describing a wide range of processing parameters are provide. ‘Since the old adage “picture is worth thousand words” ie expecially apt ina discussion of scsmie dato processing, figures make up the major poction of this textbook. In preparig some of the figures, T received reat acitance from my colleagues at Western Goo physical Company. Thanks are due to Darran Luens, Mike Cox, Greg. Godin, Dave Nichols, Tania Backus, ‘Tomaso Gabriel, Dave Hil, and Raphael Tortotawx Seismie Data Analysis “Thanks also are due to the oll companies and contrac- tors for supplying data ad some figures for which spe- ‘ic acknowledges me in the figure captions. 1 tesprest my deep appreciation to: Soraya Brombacher, Mark Wilson, Wayne Johnson, Mike Jungaieke, and Pam Jakubowies for the artwork on most ofthe figures Talo extend my appreciation 19 Meg LaVergne, who put the final touches on many fxgures end compnter- Grafted the flow diagrins. Thanks ako to the mea bers ofthe playbaele group at Western's London Dige ital Centers Stephen Blick, John Byrno, Mike Rye, ‘Chris Godave, Steven Grace, and Tony Levent. Sally Humphreys, Jan AMiteell, and Vivian Milson helped lay the tent into the word process Tacknowlege with great appreciation the review work doue by Jon Claerbout, Sven Tretel, Jolin Sher- wood, Fred Hilterman, and Greg Godkin. Yao thank Ue fellowing individuals who partielpated In review ing parts of the eater drafts: Roo Chambers, Aftab ‘Alas, Bruce Cassell, Kar! Millan, Tony Kort, Dave Brown, Darko Tufekele, Pete Bibhy, Jon. Ferguson, ‘Mark Dope, Weudell Wiggins, Jeff Resnick, Walk Lay, Bill Dragooet, Mai Yang, Patsck Ng, Stewe Cole, Larry Scott, Ken Larner, nd Holmat Jakubowicz. Special thanks are de to Diane Parker, who dd an outstanding Jol of editing fr style, proofreading. and preparing the famecirrey copy of the book. Thanks to Lyan Gui fin for helping to bring the text up to the SEG stan- dards of ptbication, Spola thanks aso are dive to my ‘editor, Stove Doherty, for his exelent and ecmpreien- Sve review. also appreciate his valuable recommenda- tons in revising the text and the help he provided in bringing earity to the text, Textend my deepest and ‘wholehearted appreciation to my wife, Hulya, for her ‘yerasting encouragement. And finally thanks are due to Western Geophysical Company of America. for the ‘support provided to me in writing this textbook. = Yilmaz London, Jay, 1987.PREFACE, ‘The frst elton, entitled Seismic Data Processing was published fn 1987 by dhe Socleey of Exploration Coo plpsicsts. Thereafter, [began to work on the second tition almost immediately. My objective ws to cap- ture cootinuovsly the new developments that were tak- ing place in the seismic industsy. The second edo js the culmination of this continons update over the past ten years The updating process was based on ex Ihaustive modol-and readata experiments with the = sults ofthe reeareh and development work of my ow ‘end many others. [have abo drawn an extensive and demonstrative st of realdata examples from the mie merous case stuies that Teondcted during the couse ‘of the update, Another source of update was of course the prolife literature on exploration ssismolony. his sccond edition embodies the broad scope of sriamic data analysis — processing inversion, and inter- pretation of seismic data. I shall give brief summary of the mast imporan new developments in sekunle data analysis dusing the past 15 years. To begin with, the {ED seismic method took a centrally dominnnt position in the exploration and development ofl and gas fields, Algorithms for 3D seismic data processing, including 4D dipsmoveout correction, 3D reaction aud tesa ‘tates corretions, and 3:D migration have now become fn Integtal part of the applications lary of the ses- ‘mie data processing systems in use today. Adaltionaly, noipe attenuation based on prodiction filtering is now ‘applied romtinely to seismic data. Techniques for m= ple attenuation based on te Radon transform and wave extrapolation have been siceesflly demonstrated on field daca. ‘Shortly after 8-D migration, we also began to image the subsurface before stacking, Eicken workflows for {KD prestack time migration are in use today not only to image the subsurface more accurately in the presence ‘of conlicting dips with diferent stacking velocities but ‘ako to generate common-rellection point gathers that ‘can be used to perform prestack amplitde inversion and ts obtain attributes associated with amplitude ‘variations with offset. 3-D prestack time migration also paves the way for estimating a $-D rms velelty ld ‘hat ean be used to perform Dix eonversion and thus ‘obtain a 3. interval weoeity fl CConeurrent with prestck maging we eran to m= age the subsurface also in depth to acconnt for strong lateral velocity variations. During the last decade, vars of effort in roeteh aud development conducted In pre- vious decades have led to practical inversion methods for earth modeling end imaging in depth. Using appro- inte inversion methods, we derive m seine reese {ation ofan carth mode! in depth, deseribod by two sets of parameters — layer velocities and rellector geome tries, for low-raie, complex, ad eomplex overburden structures. The power of -D visualization has given us the ability ta ereate an earth model in depth with the fccuracy needed to image in depth, and that within an fleet work schedule. Additionally, the rapid growth in computer power has enabled us to generate an earth Jnnage in depth from 3-D prestack depth migration of large data volumes, again within acceptable work ceh- tes, ‘To get the most out of the image volumes derives fiom &D prestack time nnd depth migrations, we now ale extensive wse of $-D visualization ia seme i terpretation. Using a volume-basod interpretation strat fey, not only’ do we pk time or depth horizons to de- Iinente the sractral model of the subsirface, but we also mnke use of the seismic amplitudes to infer the Aepestional model ofthe subsurface. ‘Tho road shad for exploration seismology ineludos ‘Hees main topies — 4D seismic method, 4-0 seismic meth, and anisotropy, all aimed at sesmie character- faation of oil and gas reservoirs andl eventually moni= Aoring their depletions. By recording 3D seismic dat, lover the fel that is being developed and produced st Appropriate tine intervals, we may detect changes fn the reservoir conditions, such a uid saturation andpore pressure, Such changes my’ be related to changes in the scamic amplitudes from one SD survey to the next, Time-lapse SD scianie monitoring of reservoirs fs referred to as the 4D seismic method. The fourth Aimeasion represents the calendar tne aver which the reservoir is being monitored. Potential applications of the 4D seismic method include monitoring the spatial textent of tho steam front following in-situ combustion ‘orstean injction used for thermal recovery, monitoring the spatial extent of the Injoctel waterfront used for secondary recovery, imaging bypaed cil, determining flow properties of sealing or leaking fults, and detect- ing changes in oil-water contact, Somme reservoirs can be klentilied ond monitored ‘ecter by using shear-wave data. For instance, acoustic impedance contrast atthe topeservoir boudary may ‘be too small to detect, whereas shent-wave Impelance ‘eontrast may’ be sufficiently lange to detect, By record ing multicomponent data at the acean bottom, P-wave tnd S-wave images ean be deeived. Commonly four ‘lata components are recorded — the pressare wase-eld tnd lnlie, crstine, and vertical eamponents of par- ticle velocity, Thus, the multicomponent seismic data recording and analysis is often refered to ns the 4 seismic method. Poteatia) applications of the -C seisiie method inchide imaging beneath gas plumes, Sal domes, and basalts, delineating reservoir bounde aries with a higher S-wave impedance contrast than P-wnve impedance coutast, diferentating sand from shal, deteetion of hid phase change from o-bearing to ‘vater-benring sands, detection of vertical fracture orien- Inn, mang dct tration, and mapping ntl recently, exploration seismology at large as ‘heen based on the assumption ofan isotropic medium, albeit we have heen coglzant of anisotropic behavior “of reservoir rocks, Seismic anisotropy often i associated ‘with directional variations in velocities Por instance, in A vertically fractured limestone reservoir, velocity in the Fracture direction fs lamer than velocity in the diretion perpen tothe plane of facturing, giving rise to ‘ethal anisotropy. Another directional variation of ‘eloeities involves horizontal layering and fracturing of rocks parallel to the layering In this cae, veloeityin the herizonta direction is higher thon the vertical dieetion, ng rg to transverse isotropy Tn addition to a continuing effort to improve the ‘esting -D time: and dept-domaln applications, eur- rent research and development in seamie data. analysis is focused on tne and depth-domain analysés of 4D tind 4 seis dat while accouating for anisotcop: ‘Topies inthis book are arganlzed to reflect the n= creasing degre of complexity in the data analysis ant the progres made in exploration seimology. Volume Seismie Data Analysis 1 is devoted to 2D conventional procesing bass on the three principle processes — deconvolution, stack- ‘ng, ad migration. Volume I is devoted to topics be yond 2.D conventional processing — &-D seismic explo- ‘ation, seismic inversion for earth modeling and imag- ng in depth, 4D seisme method, 46C and anisotropy. Fach chapter fs accompanied by an ap- ens that inchides a mathematiesl treatise of selected topes from the chapter itself. As such, practinl aspects cof scsmnic data analysis are tveated within the chapters tHhtselveswithont the burden of the theoretical tals ‘When wood as a textbook in a university, T recom mend Volume I for a frstsemester senior-level ours ‘and Volume TI fora second semester seufr-lovel course ‘ora first-year graduate course. Optionally, your may eon Siler on additonal one-semsster senior ot graduate level course on the applet theoty of exploration seis- ology sed on primarily the appendixes Thyou ave seaale analyst using thi reference, you can strdy the practical aspects of set- tie data analyst in roation to the projects you are fumdueting to get helpful hints on the algorithms and workflows If yon ate a research geophysics using this book asa ruferce, you can study the practical aspects of a specie applieation of interest to get helpfil hints fon what assumptions ean be uaade iu relation to that fanplicaion. Ako, you can study the appendixes to initi- ‘te yourself into the base theory on the subject of your interest. have a passion forthe sesnie method that T have malotained throughout say earcer. While the souree of this passion is indisputably my teacher and Heong friend, Jon Claerbont, I have been very fortunate to Ihave worked with some of the most talented individ Als who have foeled my enthusiasm for exploration sei mology. Most appropriately, T wish to express here my heartfl gratitude to each of these individuals ‘To begin with, Lam deeply grateful to Steve Do- berty, my technical editor and Belong frend, for his incisive, meticulous, and prompt eitng of this eatire work, Steve's editing brought clarity and precision to the tex. He was also the eto ofthe frst edition; on that occasion and now, he gracefully shared the experie fence with me. Thank you, Steve, fr your dedication and fffort, A part of the Cechmiea! editing, I also recived feat asistance from Zhiming Li, who edited Appen- ‘ices A through J, Joo Stefans who edited Appendix Land John Toldi who sited Chapter 11. T thank all fof you wholeheartedly for your eneful editing of the test, debugging the equations, and introducing clarity to derivations Next, [ wish to express my sincere gratitude to Judy Hastings, ny technieal copy editor, for her in presively consistent eiting of the entire mamscript book asaPreface sit ‘and he diligence in keeping me on schedule. Thank yon, ‘ord, ako for your graceful handling of my frustrations ‘with thes and as, Cwish to extend my thanks to Ted Balanjian, Publications Manager of SEG, for his sup- Port anu excellent coordination of the tasks involved in the publication proces. "Now T wish to extend my special thanks to Cyril Gregors, Ifan Tauritanir, Ferudnn Kile, Osan Yib maz, Fugen Zhou, Ma Xae Ling, and Huseyin Ozdemir, ‘who lilped me with several of the cave stuies, and model and retbdata experiments, Speefeally, Cyst hoped me erste the case studios for my 1900 SEG Di ‘Unguished Lecture Tour: these cas studs are ineluded jn Chapter 10. Cyxl has been involved in uany of the projects over the past ten years associated with this ‘work, Irfan, Ferudun, and Pigen all worked with me on twee ease stules whieh are ako eluded In Chapter 10. Orhan did the procising of the 4C data presented Jn Chapter 11. [eel very fortunate to have had such ‘exceptoually talented individuals helping me to bear the burden, Tish to extend my special thanks to David Lum- ley, 4th Wave Imaging, and Chevron for contetbuting ‘examples to the seetion on 4D seismic method. I wish to express my since appreciation to Mase Reshe for ‘renting some of the synthetic data sets I used in my ‘experiments deseribed in Chapters 8 and 8, Likewise, T ‘extend my sincere gratitude to Bygeay Landa for ere- ting the synthetic data sets I used in my experiments exer in Chapter 0, Extending the ist Tepes ry thanks to Rd Crase, Chris Taylor, Dave Nichols, Di ‘ane Dopkin, Gerald Kidd, Rob Blond, Cerys Bianeardi, Davnd Babave., and Lee Bell for providing examples for helping me create them. Laan sure that In unable torecall many ofthe names asociated with the update sgolng back tem years; ¥ thank all of you most sineoraly Tam very grateful to Chevron, Mobil, Britannia, ‘Talisman, Husky Oil, BP-Asmoco, Shell, AGIP, Total, BHP, Gulf Canada, ONGC, Sheuglt Oil Field of CNPC, Sout Aramco, and many ether eompanies who may oF ‘ay’ not have preferred to be anonymous for providing field data that enabled me to condict my experiments: LL wish to sineoely thank Damie Skeet and West- rn Geoplyseal, Rutt Bridges and Landmark, Sehhun- Derger Geeo-Prakla, Walt Lynn and PGS, and expe- cially Edad Weiss snd Paradigm Geophysical for their ‘most imaluable support and encouragement in my ‘quest fo complete this work "Auxd now the lst word, but a special tribute to ny wife, Hlyo, and ey son, Ben. Fam wholeheartedly sratefl to you both for your ending love and supa. ‘This woek undoubtedly demanded sacrifies and it was Aefinively a sacrifice on your part. You demanded very lide of me, and you gave the whole of your love to me. shall ees i all my lie 62 Yilmaz London, May, 2000.INTRODUCTION «+ Processing of Seismic Data # Inversion of Seismic Data Interpretation of Seismie Dat ' From Seisnie Exploration to Seismie Monitoring ‘The Classical Greeks had a lve for wisdom. “ecame down to ws as phil sophia. ‘And Uhave a passion forthe seismic method — Let this be an oe to philosetamos O uno scot itis Listening tothe ecs from the earth, ‘The seine method has three principal applications: (4) Delineation of near-surface geology fr engineeing ‘studies, and coal and mineral exploration within & Alepth of up to 1 km: The sosmie method applied to the near-surface studies is known as engineering sciamolony. (b) Hydreenchon exploration and development within ‘depth of up to 10 kan: The else meth! applic to the exploration and development of oil ad gas fields is now as exploration seismology (6) Investigation ofthe earth's crustal strvcture within ‘depth of up to 100 km: The seismic method pplied to the erustal and carthqnake studies is Known as earthquake setsotony ‘This book is devoted to application of the rellction senmie metiod to the exploration and development of ‘il ae gas fis ‘Conventional processing of reflection seismic data Yds an earth ine represented by a seismic section ‘which vsually is displayed in tine. Figure FL shows a selane rection from the Gulf of Mexico, nearly 40 kan fn lenge. Approximate depth scale indicates a seinen tary section of interbedded! sands nnd shales down to 8 fm. Note from this earth Suge © salt sill embeded inthe sdimentary sequence. This alloethonons sat sll thas a rugose top and a relatively smooth base. Note the folding and faulting ofthe sadimentacy section above the salt. “The reflection seismic method bas been used to de- lineate nearsuriace geology forthe purpose of coal and mineral exploration and engineering studs, especialy In recent years with increasing aovetanee. Figure I-20 shows a seatie section along a 50D-m traverse across ‘hadron valley with steep flanks. ‘The ithologie cok ‘umn based on borehole data indicates a sedimentary feczuence of clay, sand, cl gravel deposited within the valley. The bedrock & approximately 15 m below the suriace at the iages ofthe valley and 65 m below the Surface at the bottom of the sales: The strong rete tom nthe seimont- bedrock boundary i. rest of the contrast between the low-vloety sediments ahowe and the high-velocity Precambrian quartz pegmatite below "The refletion seismic aaethod also has Deen sed to delineate the ervstal structure down to the MohoSeismic Data Analysis seer) “Hopmioydsa Ho amoxp pe ope (sou paw wore388 8 8B ss (0) A shallow sefleetion seine section from Ontario (Pullen ane Hunter, 1980), and (b) dep reflection sexe rm sntnast Turkey (Vilma, 197).4 Seismic Data Analysis Alsontinulty and below. Figure £2) shows solame Section recorded on land slong &15-km traverse. Based ‘on regional control, es known cha the section consists ‘of sediments down to about 1 km. The reflection event at 6557 6, which corresponds toa depth range of 15-20 km, can be postulated the crystalline basement. The group of rflections between £10 s, which corresponds fo a depth range of 25-35 kin, represents transition zone in the lower erust — mest likly, the Mebo discon tinuiy, isl Commen-midpeint (CMP) recording ts the most widely nsed seme data nequistion tocique. By poe viding redundaey, measur as the fold of coverage ln the seisnie experimen, It inprones signal alts Figure [3 shows sexnie data colleeted along the same traverse in 1965 with sngle-fold eoverage and i 1995 vei twelve fle coverage. These to diferent vintages of data have been subjected to diffrent treatments i processing; nevertheless, the fold of coverage haa eased the mast difference in the sgual lew of the final soe Selamle data processing strateges and results are strongly affected by field acquisition parameters. Ad- Aiionlly, surface conditions have a signifies impact fon the quality of dats collected in the field. Part of the soismie setion shown in Figure Fl tween mi points A and Bis over an area cover with karti imo stone. Note the continuo reflections between 2 and 3 ‘outside the inestone Jhereney semblance spectrum (Section 9:3) to determine the interval velocity profile, you make a judgment as to what degree of lateral veeity variations needs (0 bhe honored Tlsese are bat two examples of Interpre tive work involved in processing and inversion of ele data, respectively ‘What is known as traditional seme interpretIntroduction c FIG. 1-17, Layer-by-layer estimation of an earth model in de See text f 1920 Seismic Data Analysis FIG. 1-18, Estnnted velocities for layer represented by thIntroduetion 22 Seismic Data Analysis ion, however; nls picking a rection time sur fico associated with a layer boundary from time rigrated volume of data or a reflector fan a dept nigrated volume of data to determine the structure map, for that layer boundary (Figure 119), The power of 3D visualization of image volumes, velocity volumes, land attribute volumes, such as those associated with AVO malysis and ncoustieimpelacceeetiaation, have dramatically changed the way selmic interpretation is done now. Interperetation no longer i picking travel tins to determine che stewetiral seo ofthe aren of Intest, but also involves manipulation of amples contained in the data volumes to derive information bout the depositional enviroment, depositional 36. ‘quence boundaries, and che internal constitution ofthe sequence units themselves. Interpretation of 3D seismic data is eovered in Section 75, while further eeamples are provided with the eae stos in Sections 1038 ad 09. ‘From Seisinic Exploration to Seismic Monitoring ‘The seismic industry has been impresively dyanic tnd creative during its 60-year history. Although i is relatively a sinall sector within theo and gs industey at large, ft has made the most sgniicant imp om i creasing proven reserves and reerveproduction Fats worlwide, ‘We sal now sketch a belo historiography of the ‘sciamie industry hefore we lak aid, The evolution of ‘te seismic industry ean be deseribed rilly in decals of development and forward leaps from one theme to another as outlined in Table H1 In the 1960s, the digital revolution profoundly umged seimic acquisition, We were then able to eoart! more data hy increasing the mumnber of chan. nels an fod of coverage. The distal revolution brought bout the noed tose digital computers to aaa the recorded data. That esme about in the 1970s when we switched from calculators to computers. Many of the data processing algorithms, including. deconvolution, ‘Table 1-1. The milestones in the seine industry 1960 rm enalay to digital 1970 rom eaiewlators to computers 19805 From 2D to 3-D 19005 From time to depth 2000 From 5D to -D From -D to 1-0 Brom isotopy to anisotropy ‘velocity analysis, refraction, and resid statis comrec- ‘ons, notmabinoveout corection and stacking, ad even migration, were implement in those years, The com ‘ter before the seventies gas person ting the eal celator now the computer is « machine and the person came the sismie analyst Tn the 1980, theses indostry vook another big step forward; it was now begining to provide the il and gs industry with 3-D images of the subsarface We need only to examine the glob reserve-produetion ‘curves over the past decades to see that the 3-D rev- ‘kton gave a big jump from 35 to 45 years for ol and ‘om 50 to 5 yeas for gas. The seismicity was a fend pushing the computer industry to the Tinie with its need for power to handle Iargmacale data volumes sequined by ED surveys ‘lua, nthe 1900s, the seismic industey was e2- able of providing the ol and ge industry with images ofthe subsurfce, not justin 2D, but als in depth Tt took years of exiastive experimental esearch to test and fleld-prove numerous methods to accurately eat ‘mate an earth model in depth and nse ito efficiently eeate an earth image in depth. Once again, the sie mic industry las cllenge the computer industry to provide eost-ffctivo solutions for mimereally intensive applications with large inpatont put operations, suc as 3D prestack depth migration, As the seismic industry made one breakthrough af ter another during its history, allo created new eal lenges for itself. Now we record not just P-wave but abso converted Saves fora wide range of objectives. Using the multicomponent seistaie method, eommonly ‘now as the 4C seismic method, we are now able tose tHnongh gas plumes eausel by the reservoir belom: Wee srw abl to sometimes beter image the subsalt ad sub Tnsalt targets with the 4-C slainie method, Using the converte Scwaves, we are able to detect the oll-water cottaet, and the top or base of the rewevoir nit th ‘we somtimes could not delineate ting only P-waves ‘We even go further now and attempt to identify tid {types in reservoir rocks discriminate sand fom shale, ant map hydrocarbon satoration, agin using the 4-6 seismic method. Our ultimate objective isto nse the seismic method, in ad »»rrmrrernne > cece FIG. 1.1-11, Summation ofa discrete mabe of somos ‘vith no phase, but wth the same po sept, seh 1 ba ted pine wavelet rpresnted by the trace fu te right (denoted yan aster). This sw nero a antisymmetric wavelet, The 180edegree phase shift changes the polarity of the aero-phuse wavelet. The 27)-lgseo plano shift changes the polarity of the zero phase wavelet, while converting i¢ {0 an auticymmetie wale. Filly, the 360egree phase shift retains the shape of the orginal wavelet. A aotaat phase sift to changes the shape of worelet. fu particular, a 90 degree phase shift converts a aymmetnic wavelet (0 an ‘antisymmetric wavelet, hile @ £80-degrec phase shift changes its polarity Figure 11-16 shows a portion of a seismic section with the application of diferent degrees of constant phase rotation, Nove the change In the wavelet char acter of the significant reletions. This difference wavelet charneter has aw impact on picking events fr Intespretation. When comparing displays of two dif ‘nt sections, which may be associate! with wo ferent intages of processing of the sae data two different Hines from the ste survey, the wavelet charter of the reflection event that is bein picked must be coo Slntent from one sectou to nother. A common mistake Tine (e) PIG. 1.1-12, The same simesidal components a6 In Fig- tre 1-1, hut with-a =O = constane-ine dey. When Samael, thew snus ykl a bundsinit symmetric Tyran ater). Figure 1-1, excep tine sit elated ‘ri the smced Cs is displaying two sections to be compared with oppo- site polarity. Tae polarity convention set by the Sox ty of Exploration Geoplscits is base on a negative twater-bottom reflection coefficient, which corresponds 10-8 positive polatits: So at, no Basie phase spectra have been examined Ainear and constant phase shifs. We now examine thee combined eet. The phase specteun i defined by ‘funetion ¢ = gy +42, where op isthe constant phase Shift ond a isthe slope ofthe linea paso shit. Figure L117 shows the result of applying a linear phase sit (as in Figure 11-12) plus a 90-legroe eonstast phase shift (as in Figure L-1d) to the ssoids in Figure 1.I-L. The zero-phase wavelet withthe sme amplitude spectra that in Figure 11-1 yas shifted in time by “112 s becase ofthe near pliwe shift, and converted to an autisyminetsie form bees ofthe constant 90- dlegece pase shift,30 Solsmie Data Analysis FIG. 11-18, Storing with a zeo-phnse walt (a), car pve aie are applied to shift the wmv a sno without ‘hang ts shape. The shape of the lncar phe anti erat tthe tne si variations in phage spectrum are shown in Figure 1.-18. The zero-phase wavelet (Figute 1.1188) can be modified to diferent shapes simply by chang, ing the phase spectrum. Tt ean be medi to the ex- tent that it may no lougor resemble the original wavelet slp as itustrated by the last example (Figure L.1- 1a), By herping the amphitude spectrum unchenged, the toeelet shape cau be changed by modifying the phase secirom. ‘Time-Domain Operations Consider asflertivty sequence represented by the time series (1,0,4), Also consider at inipalsive source that catses an explosion st ¢ = 0 with an amplitude of 1 ‘The response of the refetvity sequence to aa inpulse ical the impulse response, This piel proces can De described ain Table I ‘Table 1-1, Response of the retectivity. sequence (1,0,4) toa zero-dela explosive imple (10). Time of Reteti Onset Sequence Sonree Response PIG. 11-14. The sane satscidal components usin Fiqute 141i but witha constant degree phase ak apple! ch Tho aero eros ae ‘Sermon of se sims yields an antisymmetric wavelet thas fe rep AOE Poa) aso feet hy the trace othe ig sented by an serFundamentals of Signal Processing. aT @) FIG. 11-15, Starting wit the zer-phase wavelet (a), its shape i changed by applying constant phase shifts. A O-degree ‘hice sft ouverts the zro-phase wavelet to an ati’ munetie wavelet (0), while a 18D-egtee past hit revert polarity {o), A 27elogreephas sit rovers the polar, whi making the Waveht naeymanetrc(d). Finally, a 360-dogro phase hit doesnot mei the wavelet () FIG, 1.1.16. A portion ofa sismiesetion with difrnt degrees of constant phase rotations.a8 Selamie Data Analysis PIG. 1.1-17. Anas (as fn Figure 11-12) combined witha ‘oust pase shit (as Figie 1-14) rvs 8 ine shied artlsymetrie welt The wavelet ls reposted by the trceom the ight (devoted Ly an astra). (@ (b) Table 1-2, Response of the reflectivity. sequence (1.0, 4) toa unit-delay implosive impue (0, —}). Time of Reflectivity Onset Sequence Response One unit time Inter, suppose that the impulsive ounce generates an implesion with an amplitude of 4 This response is deserbed in Table Note thatthe response in each ease the reflectiv= Sty eqence sofa by the impulse strength and delayed by the impnise onset, Since general src function considered to he a saquenee of explosive a implasive impulses, the individual impulse responses are added to obtain the combined response. This proces is eld linear superposition and is deseribed in ‘Table 13 Convolution, In Table 13, the asterisk denotes coneutution. The re- sponse ofthe rfetivity sence (1,0, 4) to the souree wavelet (1,3) was obtained by convolving the two series, This fs done computationally aa shown in Tie am FIG. 14-18, The shape of « xeo-plivo wavelet (a) ean be modified by inte & nonsco-plase spectrum of any formFundamentals of Signal Processing, 30 ‘Table 1-3. Linear superposition of the two responses deere in Til 1-1 nnd 1-2 Tine of Reflectivity Onset — Sequencer Souree Response ° roy 1 10} Superpeniton: Bipressed dferently: (1,0, 9) (= Blo 1-4 A fixe array bs set up from the reflectivity sequence. The source wavelet is eversed (folded) and moved (lag!) one sample at atime. At each lag, the el- fees that align are malipied and the resulting prot ets are same ‘The mechanics of convlution are described in Ta Be 1-5. The number of elements of output array cx is given by mtn 1, whore m andr the length of the ‘operand array a; and the operator array By, respectively: When the roles of the arrays in Table 1-1 are in ferchanged, the output array jn Table 1-6 results. Note ‘that the output response is identical to that in Table ek. Hone, convoletion ss commutative — i does rot ‘matter ehich array is fd and which ix mowed, the out ul is the some. Crosscorrelation and Autocorrelation Seismic data processing often requires measurement of ‘tho similarity or time aiauneut of two traces. Corela- ‘able 1-4, Convolution of the source wavelet (1,4) ‘ith the reetivity sequence (1,0, 3). Rallestivey Ont Soqence Response 10 ae 1 ae i ct ‘Table 1-5. Mechanics of the convolutional process. Fixed Array Moving Array: bide Given two arrays a, and by Step 1 Reverse moving nee by Step 2: Multiply in the vertical diroetion. Step 3; Add the products and write as output ey Step 4: Shift area B, one sample wo the eight ‘nd repeat Stopa 2 and 3 Convolution Table 013054 5 ty Ont babs by « bb by a bah by a babi by « babs by a bint s bab be & bhi by = bg by bo & a by by o where ce Danby AOL mena1 ‘Table 1-6. Convoltion of the refctvity sequence (10,4) withthe source wavelet (1,5). Souvee neput Wavelet Response 1-4 poo. 1 boo yoo t40 Seismic Data Analysts CCrosoortlation of wavelet 1 with wavlet tion is another time-domain operation that is used to alee sich measurements, Consider the following two wavelets Wavelet 1: (2,1,—1,0,0) Wavelet 2 (0:0,2,1,—1) Although these wavelets are identical in shape, wavelet 2 i shifted by two samples with respect to wavelet ‘The time lag at which they are most similar ean be de- ‘termined To do this, perform the operation on wavelet 1's described in Table 1-5 sithout reversing wavelet 2 (omit Step 1). This is eroscorrelation and the result is shown in ‘Table 1-7. Croscorrelation measures how such two time series resemble each other. From Table 1-7, aote that maximum correlation fceurs nt lag -2. This suggests thnt if wavelet 2 tere shifted two samples back in tine, then these two Trovelets would have maximuan similarity, “Table 1.8 shows the exoscorelation vals that re- sult when the atrays are interchanged. This time the ‘maximum correlation occurs at leg 2 Thos, if wavelet ‘Table 1-8, Crosscorelation of wavelet 2 with wovelet ‘Output Lag 0-4 0-3 0 ot 20 1 62 13 24 ‘Table 1-9, Cromscorelation of wavelet 1 with wavelet 2 and erossorrlaton of wavelet 2 with wavelet 1 Grosscorelation Wavelet 2 vee 1 with Wavelet 2. Owtpnt Lag with Wavelet 1 2 “4 o 1 =3 ° 6 = o 1 e 0 4 ° es 0 1 1 ° 2 6 0 Z 1 0 4 2 1 were shifted yy tro samples forward in the, these to wavelets wold have minum silat. like convolution, eroscorrelation $8 not comme {ative — the output depends on hich array te fixed and ‘ehich is moved Table 1-9 shows comparison of the rosscorrlaton results Usted in Tables 1-7 and 18. ‘Cromsortelation of time series with itself known 1s avtocorrelation. Table 1-10 shows the antocorelation Ings of wavelet 1. Note that maximum correlation oc crs at zero lng, an Important property of autocorelar ‘lon. Moreover, the autocorrelation function is ymmet= re, This & property ofa eal tne series. Therefore ‘only one side of the autocorrelation needs to be com puted 1 ie heuristically shown in the next soction on fre- ‘qvency filtering that convolution in the time domain is ‘equivalent to multiplication in the feqeney domain (Bracewell, 1965). Since corteation is a convolution withont reversing the moving aay (Table 1-5), a sh ‘Table 1-10, Autocorrelation of wavelet | 21-100 ‘Output Lag 21-100 4 21-100 0-3 21-100 2-2 21-100 ao 21-100 Boo 21-100 oe 21-100 22 21-1 00 0 3 21-100 0 4Fundamentals of Signal Processing a bette cots perce rere Pete “is ea Bee see rere oe sae —! mero Ont ance —— es wien ‘eam FIG. 11-19. Tho frequency-domain description of conve- fetion and eoreaion. Jar frequency-domain operation alko applies to correla tion Figure 11-19 is « summary of frequency-domain descriptions of eonsolution and correlation. Note the Following properties for couvolation and coreat ion (9) Both convolution snd eorelation produce an out- ppt with a spectral bandwidth that is common to Doth of the input series. Th immediate example is tho band-pass filtering proces. (0) Phases are adltve incase of convolution and sub- tractive in case of correlation (Bracewell, 1965). (e) For autocorrelation, this implies that the outa series is zoro-phase. This fact already ws verified by the example in Table 1-10 where ft was shown ‘hat the autocorrelation i symmetric with respect to nro lng As a measre of similarity, eresorrelation is used Widely at vations stages of dlnta processing, Tor i tame, traces in @ CMP gather are eresscorreate with 2 pilot trace to compute residual staties shifts (See tion 3.3), The fundamental basis for computing veloc- lity spectra i crossoration (Section 32). The con- sient elements of the Wiener filter (Section 2.3) are ‘reasconclation of the desired output waveform with the inpnt wavelet, and antocorrlation of the iupnt wavelet Vibroseis Correlation xe other important proces isthe vibrosciscorelation, ‘This involves croscacelation of suey sigal with the recorded wibrasels trace. The surep i frequency rovllated vibes source sigual input #0 the ground ‘The convolutional model for vibroseis data is desribed in Section 25 igure 11-20 shows a wibroseis sweep signal, a recordei| common-souree gather, and the eorrclated gather. The sweep length i 10s witha frequency band Of 6 to GD Hz. The 15s uncorrelated vibrosels sword Yields 6-8 coreoted record. Note that the ently part ‘of the uneorelated record contains low-frecency ray with inerosingly higher frequencies at late times ‘This ig beeatwe an epsweep (frequency increasing with ‘me) signal was aod inthis data example, Frequency Filtering What happens to a wavelet when its amplitude spec- trum is changed while its zero-phase character is pre= serve? Ta begin, consider the wavelet in Figure 1.121 (summed trace 1) resulting from superposition of two ‘very low-frequency components, Then, avd increasingly higher freqeney components to the Fourie synthesis (summed traces 2 throngh 5). Note thatthe wavelet ‘the time domein is compres as the froquency band- width (Ihe range of frequencies summed) is increased, Ultimately, i all the frequencies in the inverse Fourier transformation are inehided, then the realting wavelet becomes a spikes cen in Figure 1.122 (summed trace 6). Therefore, a spike ix characterized as the in-phase synthesis of all frequencies from zero to the Nyquist. For all frequencies, the amplitude spectrum of a spike is unity, while its phase spectrum is zo. Figure 11-23 shows five xero-pase wavelet, sy thesized a6 shown in Figure 11-21. Note that all of ‘them have bond-linited amplitude spectra. A zero- phase band-limited wavelet ean be used to filter seis Imie trace. The output trace contains only those fre- {quencies that make up the wavelot used in filtering The Une-domain representation of the wavelet is the fier operator. Te slividal tne samples af this operator ‘ate the filter cveficient. The process described here i ‘2ono-phaso frequency filtering, since it docs not mod- ity the pase spectrum of the input trace, but merely ‘bandlimits its amplitude speetram, Frequeney-domain filtering involves multiplying the amplitude spectrum of the Input seis trace by that ofthe Blter operator. ‘The procedure is described in Figute 1.1-21, On the other hand, the Sltering pro- ‘cas in the time domain involves convolving the filter ‘operator with the input ine series. Figure 11-25 is a ‘deseripton ofthe flter design and its time-domain ap- plication. The frequeney- and time-domain formulations ‘of the Bering process (Figures 11-24 and 11-25) are bated on the fullowing important concept in tne seriesa Seismic Data Analysis ° 2i7 3b 68 abo 5 10 HESS 19 BEEN Sweep " Uncorrelated Vibroseis Record (6-60 Hz) Correlated Shot Record FIG. 1.1.20, Vibrowis correlation: the ence sghal i correlated withthe recorded vibes ond to get eoreatd eld data, A'I0seswoep and Ths record dain yields cordate reo analysis (Section A.1): Convolution in the time donscén te equvalent to matitintion in the frequency domatn Sinslarty, consolution in the fequency domain is equie- alent fo multiplication in the time domain. Frequency fltering ean bein the form of band-pass, bband-reeet, high-pass (Jow-eut), or low-pass (hight) filters. All of these ites are based cm the same prine ple — construction of « zexo-phase wavelet with an amplitude spectrum that meets one ofthe for speci ‘Bandpass filtering suse inoxt commonly, because «a seismic trace typically contains some lon-hecquency ois, such as ground rol, and sme highefrequency am- bient noise. The usable seine reflection energy tually 1s confined to a bandwidth of approximately 10 to 70 Hz, with « dominant Frquency atound 20 Hy Band-pass filtering i performed st various stages in data processing. Te necessary, i can be performed before deconvolution to suppress remaining grouned-rll energy and high-frequency ambieut noise that other= ‘wise would contaminate signal autocorrelation, Narrow band-pass tering may be neemeary before erosccrte- lating traces in 8 CMP gather with a pilot trace for use in estimating residual statics shifts (Section 2.)Fundamentals of Signal Processing Time (s) : 0 0 0 6 06 12 Oo Frequency (Hz) 6 12 18 24 30 E11 13-21, Tho memati of xo-phane smoke with denial peak amplitudes. Trace rest fom ech summation ire ttbered fom 1 to 8. As the fequency bandwith ¥ increntd, the ayhasond eropliae wavelet Lnredingly 6 Time (6) oat cal ee 1% 3 BOT Frequency (#2) FFIG. 1-22, The output wavelet becomes apie (ated tice 6 [to the Noes een. Compare tis with output trace thre the summation inchs snide tal rogue ‘in Figure 1-24 Seismic Data Analysis MOT aT FIG. 1.1-28, & series of or-phaso ware Is inereased, the wavelt is mane compressed in time Band-pass filtering also can be performed before com- ting eroseorelations during constevetion of the ve locity spectrum for improved velocity picking (Setion it sa standard practice to apply a time ‘variant band-pass filter to stacked data (Section 1.5}. Practical Aspects of Frequency Filtering Applicaton of filter in the frequency or tne domain (Figures 11-24 and 11-25) viekds basin identical re sults. In pesetiee the time-domain approach ie favored, since convolution involving a shart areny, sch as afk ter operator is mare economical than applying Fourier transforms, Define Desir Input Selemie Trace Amplitude Spectrum Tor the iter Fourler Transform Amplitude Phase Spectrum Spectrom aigiy —! Inver Fourier ae tered Output FIG. 1.1.24. Design and application of 0 zerw-phase Ser Jn the fegueney domain, (Cop row) and thee respective ample pectin (bottom rom). As bandit 23, the fundamental property of ‘tated as follows: The bronder the baci, the more compressed the filter operasor: tas, fewer filter cooficients are required. This prop- erty also follows from the fundamental concept tat the time span of o time series 4 inversely proporionel fo its spectral banda Tn designing a band-pass iter. Ue goal isto pase a certain bandwidth with bile or no modification, and ‘to largely suppres the remaining pert ofthe spectrin fas much as practical. At fist, it appears that this goat an be met by defining the desired amplitude spectrum for the filter operator as fllows: wna{y hats a-3) Define Desired Amplitude Spectrum Set the Phase ‘Spectrum Zero Inverse Fourier “Teamtorm t er Operator { Input Seismic —~ Convolve “Trace 1 Fitered Output FIG. 1.1.25, Design of ero phe Goqucney Mor sa ite application in the tne demalnFundamentals of Signal Processing 45 enc stune @ (co) FIG. 1.1.20. Three mo-phase wavelets (Wop so) ane their repective ampli spectra (Doom row)- (a) The serpy ‘ined ope of the pans eee pos the wavelet sed he acts spine spect (0) A sdetae and (e) gale ope help eliinate the ripples. Refr tothe text for discon of corner Requeres A,B,C, aad D- where f: and fs are the cutoft frequencies. This is known fs the Boxcar amplitude spectrum. ‘To analyze the properties of such a filter, perform the following sequence of operations (W) Define boxoar amplitude spectrum and zero- phase spectrum, (b) Apply inverse FFT and obtain » fuer operator. {e) Trineate the operator, (@) Apply forward FFT and compare the amplitude spectrum of the truncated operatr. ‘igure 1.1264 shows the sults of this sequence of op- nitions. The operator is at the top, aud the vctual and desired (boxear) amplitude spectrn are superin- ‘posed at the bottom. Note that the actual spoctrum ‘asa ringy charcter, This known as the Gibbe phe- ‘uomenon (Bracewell, 1965), and rests from represent Inga boxcar with a faite mauber of Fourier eelicients. From a practical standpoint, the ringing it wodleseable, slice some of the frequencies in the passband are am pli, while othors axe attenuated, Additionally, some fof the feequeneles in the meet zooes on both sles of the boxcar are passed. How is the Gibbs phenomenon circumvented? In- stead of defining the desired passband as a boxcar, ‘assign slopes on both sides as shown in Figure 1.1 26, and ths define the passhand as trapenaid. Note that the actual and desired amplitude spectre are now closer in agreement and the aperitor is more compact {Gt as fewer nouzoro coufconts) However, in achiev- ing a moro compact operator, the hupo of tho dosed spectrum has been compromised and. the passband is baoader that intended. ‘The trapraold slopes must be ‘offen gentle to achieve a saisctory’ resulta in Figure 11-26, where the actual and desired spectra ate approsdinately equal and the operator ix compact. ‘This is mast desirable in practice, sine it fs best 0 ‘work with operators that areas short possible It is recommended that & gentler slope be assigned om the hhigivfrequeney side relative to the low-froqueney side of the paband. Finally, while defining the passband 18 a trapezoid, smoothing also must be applied at the46 Seismic Data Analyst FIG. 1.1.27, Sold bare indicate tho Ie lng (nonmro coetients) ofthe bad pase flee operator. Severe truncation (a) ass sigaleant departure ofthe atl a ‘same nal ie cass. ‘comer frequencies (A,B, C. and D, as indicated In Figure 1.1-26e). This must be done because the Fourier transform esists only for eoatisous functions (Bracewell, 1965) How short can the operator he? Figure 1.127 shows a sequence of increasing longer operators. Sol bars indicate the operator lengths of the truncated. fl ters Note that excesive tanention eases a Tage devie ation ft the desired amplite spoetrum even though reasonable slopes were provided to the passband. Ex: tension of the operator length brings the desced and rctual spectra closer. However, there i certain length Deyond which nearly zero culicionts are added t0 the opecetor. The erteion that used to define the ope: tor length is that the frequency bandwith os inversely Proportional tothe effective length of the operator Js intimately tied to vertical (tem- poral) resolution of seismic data. Consider the filter ‘operators in Figure 11-28, Both have the same offec- tive bandwidth — the differeuce betwen the highest and low-cut frequencies. Therefre, the envelopes of the tro operators are identical. The grenter ringyes of the secon operator (Figure 11-28b) results from ts lower ‘bandwidth ratio — the ratio of the high-ut t the low ft frequency. ‘There Is a commen misunderstanding that only high frequencies sre needed to increase temporal reo Tution. This uot true. The top frame in igure 11-29, ite spretum fom the desired (tapered) amplitude spectrin, which i the shows a single relloctor and thre sets of closely situated reflectors with 48,24, and 12-8 time separations. A of neon band-pass filers apple to thew data a shown in the lower frames, The reflectors with the 48.3 separation are resolved reasonably well by wing the 10-1020-He bandwidth, However, the more closely FIC. 1.1-28, Two wavelets (09 tom) with the same band ehh (bottom so). The passband of molt (a) nner 115 Fg, while that of ware (b)iseemered af 35 Hz Both ‘ravelts have ripples, althowph one low and the othe i igh reqaeney in character. st having ot high eee ‘des dae not wfc both are needed to cree tempt aslo,Fundamentals of Signal Processing a 08 08 08 yy 08 fl 1 1 12 120 12 08 ar 0.8 il 1 —— 12 10-20 Hz 0.8 1 12 14 1 08 ie 1 1 So-sore = 08 08 12 12 3 30-40 He 1 - 1 08 mmr 0-8 + > HINER a 12 Sa jo-sone 1 1 08 se - 08 i 12004 ‘ Et 08 12 so cone : Vgc. 1.2.20." top ston tose efectvity model 4.2. IN Fg 1.1.2 Refers th age sprain (9) 50-60 Hz FIG. 1.1-26. The top seetion isa rellcivity model tat, ‘omits of fom lf to right, three rcs with Ans sep fraion theo rls with 2m sapartion, thew elie tors with 12-18 separation and single refsctor — all en fered 1s. Band limited responses (the vase bandh, 10H, center a diferent frequencies) donot provide good reluton situated reflectors cannot be resolved wih this Alter For the 20-t0-0-1i bandwith, again, the 48-ms flee. tors are reasonably separated. Nevertheless, none of the ruarzow band-pass filters provides the resoition needed. fo distinguish the reflectors situated with smaller sep- rations. Just having lw or high frequencies dacs not tre recall wth bandwith a low ae 1080.90 He, Hom ver, reectors with smaller seperation (24 ad 12s) = ‘que incearngly largr bandwidths for rslaton. ‘improve tempore resolution. Both low and high freguen- ies are needa to increase tenoporal resoution, Tis is ‘demonstrated fnther in Figure 11-30. Note that closely ‘stnnted reflectors ean be relied only with increas ingly bronler bandwidth. The 10-to-J0-Hz bandwith fs silficient to role the reflects with 48-1 separa tion. The 10-to-50-He bandwidth i sufficient to resolve the reflectors with 24-ms separation. Finally, the 10- to-100-He bandwidth is needed o resolve the reflectors tat are separated by 12 ms, There is close rlation- ship between the amount of separation and the desined Dbandwidth (Section 11.1),48 Seismic Data Analysis ‘Time-Variant Filtoring ‘The stianie spectrum, especially the high-frequency end, is subject to alxorption along the propagation path Deeanse of the intrinsic attenuation ofthe earth (Section 1.4). Consider the portion ofn stacked section and its narrow band-pass flere panels in Pigute L131. A sig nal is preset fram tap to bottam within the 10-t0-20., 20-10-30, Ueto, and 40-4 50-Hez bands. Not much, signal Is noted below 3.5 5 im the 50.to-60-He bata "Neverthelass, the signal comtent appears tobe retained oven to 35'9 with the 60-to-T0-llz bara. Finally, the 70.40-80-He band shows signal down to2.5 5. Higher f= ‘queucy band of wef signal are confine tothe shallow part ofthe setion. Thus temporal resolution reduce svealy in the deeper portion ofthe setion, ‘From a practical standpoint the time-variant char scter ofthe signal bandwidth requires an application of frequency filters in a time-varying mnie. By so do- ing, the ambient noise, which begins to dominate the signal at late times, is exeluded and a setion with a Ihigher sigal-to-oise rata ie obtained, ‘Table 1-1 liste ‘the time-variant fer (TV) parameters selected from ‘de panels in Figure 11-31, The tered selon i shown ton the far ight panel ofthe same Figure, In practice, the filters are blended nctossaljacent time windows toe tablish a smooth transition of the passband regions ‘A second band-pass series of filter scans, which i shown in Figure 1.132, allows au assessuent of the right eholce of the bandwidth for « given time gate, Hero, we start with o narrow bani filter at the low frequency end ofthe spectrum aad gradually broaden the passband by including higher frequencies For some data, the bandwidth may be kept quite large from top to bottom, ‘The stacked section in Fig. ture 1.132 ean tolerate wide-band filtering from early to late times. The filter panels for the stacked section in Figue 1.1.38, however, indicate that the signal band rapidly becomes confined to lower Frequencies at Inte ‘times. A signal is present from tp to bottonn within the ‘Table 1-11, Time-variant filter parameter forthe data shown ia Figure 1-31. The ffequeney values core. spond to the corner fequencles daseribed in Figure L.l- 26 Time, Filtee Band, is We o 5,10-70, 80 2500) 10-60, 70 3500 5, 10-50, 60 000 5, 10-40, 50 frequency bands up to 40 Hz, Nose is noted below 2.5 ‘in the 40-o-50-Hz band, This nose quikly bales up (0 shallower tlmes at higher frequency band ‘Time-variant filters typically ae applied on stacked ata, A uniform bandwidth rust be established wen filtering two sets of data that may lave diferent vine tages, source types, or noise levels. This is expecially significant when trying to te to lines and follow a r= Actor across them. The interpreter uses the fruency ‘character ofa marker hovizon a reference inthe track- Ing procere, Therefor, two intersecting ies soul bo filtered so that the refeetion character is consistent from one tothe other this sinplifying the inerpreta- 1.2 THE 2-D FOURIER TRANSFORM: ‘Multichannel prooesing operations ean be loosely de fined as thove tat mst operate on several data traces, ‘simultancously. Multichannel procases canbe use i dliseriminnting against noise and eahancing signal on the bnss of» erlterion tht ean be distinguished from trace to trace, sudh as dip or movent, ‘The 2-D Fourier transform is a basis for oth analysis and implementa ton of multichannel proces Consider the six zeroofeet setions in Figute 1.2 1. Tho timee spacing i 25m with 24 tonces per section, All have monoehomatie events with 12-He frequency ut seth dips that sary fom 0 vo 15 my/trace. Prom the discussion on the 1-D Fourier transform (Section 12), we know about froquency, particularly tempore Frequency, oF the nuinber of cyeks per unit tie. ‘This is the Fourier dual for the tine variable. However, stismle waved is not emly a function of time, bat als 8 function of a spac variable (oft a ankdpoint, ‘axis). The Fourier dual for the space variable is defined, as spatial frequency, which i the number of eves per Init distaee or maeemumber, Just asthe temporal fe quency of a given sinusoid is determined by eonnting ‘he uumber of peaks within a unit time, say 1, the swavenninber of dipping event is determines by count- Ing the anmber of peaks within a nit distance, say 1 kn, along the horizontal diction. Just asthe feuspo- ral Nyquist requency is defined as in equation (1-1), the Nyquist wavenumber is defined as 1 ar where A is the spatial sampling interval. For all of the sections in Figres 1.21 through 1.26, te Nyquist wavenumber ft 20 eles, since the trace interval is 25m. vor a-4)Fundamentals of Signal Processing. 49 FIG, 1.191, The far let panel ia portion of CMP stack without tering, The fallowing panels show the same dats ‘ith ileent nano fdas filters. The fequency bands specified correspond tothe comer Goqvencies B and C in Figure To atl high eqweney eas of cach pecan. The far right panels the Saupe sections tnt in te far lef pel after the applleaion ofthe te variant filter specified in Table 1-1 input 10-20Hz 10-30 10-40 10-50 10-60 10-70 10-80 FIG. 1.1-92. The far left panel i & portion of a CMP stack without filtering. The remaining panels sow the same data ‘eth dient ban ps filters which have Increasingly wider passband. The fequency bands specie corespun to the ‘otter foquencies Band Cn Figure 1.26. Appropriate slopes were msg to both low and high-frequency eas of each mebanSeismic Data Analysis input. 8-12 Hz 12-16 16-20 20-30 30-40 40-50 50-60 FIG. 1.1.83. The upper fr left panel is» portion of « CMP stack without Altering. The remaining panes show the same ata wit diferent arom and-ps fiersFundamentals of Signal Processing a ‘To compnte the wavenumber thot 1 asociated ‘with the section corzesponing, sa 1 the 15-ms/trace ip in Figore 121, follow a peak or trough across the fection, Fist compute the total ime dilference along the selected! peal or trough across the scion: (28 traces) (15 ms/trice) = 345 ms. ‘Then convert this to number of eyeles by dividing by the (temporal) period! 45 ms [ioovima/s) (2a) ‘The spatial extent of tho section is 575 m; herefor, the wavenumber asociated with the 15-ms/trace dip fd the 12-Ha frequency is Aldeyeles 575k ‘To continue this discussion, we will map these sec tious to the plane of temporal frequency versus spo- tial wavenumber, then look ut (wo quadrants of this plane. Tho fllowing convention will he used: Events with downdip vo the right are assigned positive dip, ‘whike events with upelip tothe right ar assigned nee tive dip. Additionally, positive dips map onto the right ‘quadrant, which corresponds to positive waveranbers, ‘while negative dips map oato the feft quadtant, whieh ‘ormesponds to negative wavennbers ‘The plane of fesquency-svvemuanber (the f — ke plane) appears at the bottom of each stetion I Fig fie 12-1, The section with zeroip eveats maps onto.a ingle point on the frequen axis at 12 He. Zero dip is fghivalent to zero wavenumber. The magnitude of the spike corresponds to the peak amplitude of the sim Side that make up the traces inthe section. Therefore, the Jk plane actually represents the 2-D amplitude spectrum of the section inthe ¢~r domain, These data Ihave heen transferred fiom the timespace domain £0 the fegiueney-wavernntver demain. This process is de- seribed motienatienlly by the 2-D Fourier transorm, ‘There ism practical relationship between the four stibles: time-space (¢ — 2), and thelr Fourier duals froqueney-wavenuunber (J — k). Measure the ineers of the stepont in the L5-ms/trace section in Figure L2-1 hy following a peak, trough, zero crossing from trace to race. Sepont is defined asthe slope At/Az. To this fxs the inverse ofthe stepot is Ac _ 057k Bt” O3ibs Now, eampute the ratio: J __1Zesele/a E> Faeyclen/kew ‘Foon this, the inverse of the stepont measured nthe ‘= domain alone m constant phase is cal tothe ratio Moyes 12eyeles/kn, LT kes = 167 kas ‘ofthe frequen to the wavenumber associated with the Ae ark ‘Therefore whe reining fied stopout, doubling the fr ‘quency moans doubling the wavenumber ‘Note tha ll ection in Fiqare 12-1 have the same frecwency component. However fom 0 t0 15 ms/trace, the number of peaks inreases horizontally across each section, That 1s for a given frequency, higher dips are ‘assigns to higher waverumbers, as seen on the J — plots. a-» Spatial Aliasing From Figures 1.21 through 1.26, comsider the same dip components but at dlloent frequencies. Map each individual section to the f ~& plane. Nothing umsual happens until the sctlon with 13-ms/teace dip at 36 Hz is reached in Figure 12-8. Horo there is no postive dip. Tn fact, as whole, the section displays a chicekerbonrd ‘character making i dificult to determine whether the ‘ip is postive or negative ALAS Ha (Figure 12-4), the correct dip direetion Ss observe in the ist four sections, However, the fifth section, which corresponds to the I2ans/traee positive dip, shows a negative di. Therefore it Ss mapped onto the negative quadrant, which ix the wrong quadrant for this section. This dip component (12 ms/race) at this frequency (18 Ha) i spatially alfosed. In fact any dip _gronter than 12 mas/trace is spatially alias at this fre quency. In the next set of sections in Figure 1.2-5, spatial aliasing occurs at 60 Hi for a S-tas/trace dip. Spatial fliasing uot only eases mapping to the wrong quai ‘ant, but also cases mapping with the wrong dip. One fbvions example of this is mapping « L5eans/trace dip fat 60 He (Piguce 1.25) Finally, at 72H (Figure 1.26), ‘tae us/trace dip companent is on the verge of spall aliasing, Moreover, the L-ms/trace dip components Spat alias tice; it folds back to the positive-dip ‘quadrant and appears at a lower dip ‘This same analysis ean be used for the negative dip components, From Figures 1.2-1 through 12-6, note that each section asa whole was mapped onto a single point in the frequeney-wavenumber domain, Fach se- tion has assoeated unique frequeney and wayemni= ber assigned tot. These 2ero-offct setions ean be con- sidered representations of plane waves that propagnte at ‘nique angle from the vervienl nnd earry a monocro= matic signal. The waveltont is defined as the line ofSeismic Data Analysts 12-He Sinusoids i i ia i i Pn i -20 20 Wavenumiber (cycles/km) FIG. 12:1. Top row: Six gathor each containing 12-1 monofequency evens with dire dips ranging from 0 to 15 Th ele tespectve amplitude specen The The slid vertical lines ae the frcjueney exc te pot from eo sgt, then al events soap onto the postive guadant Inthe Fequeney-anrontaober Ina series ois gues that deserts napping of montequeacy signal othe Gf) mstrace =O. 3 6 9 12 15 s : -20 20 Wavenumber (cycles'km) PIG. 1.2-2, The sume as Figure 12-1, execpt slg 24Tle manotegweucy eventsFundamentals of Signal Processing msitrace 0 3 6 9 12 15 | ikl A =20 © 20 | Wavenumber (cles) FIG. 1.2-8, The same ag Figure 12-1, xcopt using 96-1 monofrequency events 48 lt 0 -20 20 Wavenumber (cyclesikm) FIG. 1.2-4. The same as Figure 1.21, except using 48-Ue monofrequency events 53Ba. Seismic Data Analysis msitrace — Q. 3 6 9 12 15 — fl | ‘Wavenumber (cycles/km) FIG. 1.2-5. Tho same os Figure 1.21, except using 60-He monofrequency events Wavenumber (cycles/km) FIG, 1.2-6. The same as Figure 12-1, except using 72-Hz monofrequency events.Fundamentals of Signal Processing 35 mstrace =O. 3 6 9 12 15 il -20 20 FIG. 1.3-7. Top row: Six gatas cath formed by sn Sci e25 m. Boctom rom: Fuspocive aampltude spectra femstant phase, while the ditestion of propagation i perpenilienlar to the vaelront. Since « slic wave fie sa superposition of mauy dips and frequencies, i Ts equivolat 1 the synthesis of many’ plane-wave cou ponents. In ehie respect, the physical meaning of the 2D Fourier transform is important, for ft isan anal ‘sis (decomposition) of waveted into its plane-wave ‘components ‘A recorded wavefield is a composite of many’ dip and frequency componets, sich as those shown in Fig ures 1.21 through 1-2-6. Suppose that setions with the se dip, but with dilfereat frequencies, are super posed. The composite seetions are shown in Figure 2-7 ‘ith the composite amplitude spectra below each sec- ton. For a given dip, all fequeney components map ‘onto the J — plane along a stmight line that passes ‘thnongh the origin, The higher the dip, the closer the rae lal ie in the fk domain i to the wavenumber ais ‘The rero-lip components map along the frequency tx rom the 8, 12., and 15-mns/trace dips, note thet the spatially aliased frequencies ne loead along the linear Stgments that wrap around to the opposite quadrant in the amplitude spectrum. The steper the dip, the lower the frentency at which spatial aliasing occurs So far, a diacrte numberof frequencies was consi ‘ed. Fora continuum of frequency components aso ‘tes with a single dip, we anticipate that they would ‘Wavenumber (cyclesikm) ng gators of the ike dipe i Figwes 1.2.1 through 126. The trace map along a straight, continuous line in the J ~ do- oain, a shown in Figure 12-8. While the dipping event n Fighre 1:28 is not alissed, the dipping event in Fi ture 12-9 apetially llasd ginning at approsimately 21 He Exainination ofthe mowochrosatie singe-dip see tons in Figures 1.21 through 1.26 shows that each se- ton maps onto a single point in the f ~ k domain. An extension of this observation is made in Figure 12-10. ‘vents with the same dip in the tr domain, regardless of ther location, map outo a single rodial ine in the fk domain, When events are spatially aliased, the tadinl line wraps around at the Nyquist wavenumber (Figure 12-11), These concepts have important practi- cal plications for they lead to f ~ k ap filtering of coburent linear nose (Section 62). Eveats with differ cnt ips that may interfere in the ¢ — domain eam be Isolated In the J — F domain, ‘The munserical computation of the 2-D Fourier transform involves two 1-D Fourier transforms. Figure 12-12 shows the steps that arc ivolved. A brief math- matical formulation of the 2-D Fourier transform is slven In Seetion A.2. Insummary,2-D Fourier transformation isa way to decompose a wevelield int its plane-wave components Each plane wave carries a raonoeheomatie signal that56 Seismic Data Analysis Ax=50m 10 o 10 Wavenumber (cycles/km) FIG. 1.2-8. A single, isolated dipping eveut and its 2 toplitde spectrum. No freuney Ss sptilly lined. What 1s the dip of the event in millinconds por tence? Propagate at a certain angle fram the vertical. Bvents With the same dip in the 1 ~ dommai, regardless of location, are mapped onto a single lie inthe radial sretion in the = domain, In Chapter 4, we shall discuss migration methods inthe J & domain, and in Chapter, we shall demonstrate use of fi ilteving to ‘remove coherent linear noise and attemiate multiples, Spatial aliasing has serious effets on the perfor ‘mance of multichannel procests sich as f—& titering (Section 62) and migration (Setion 4.6). Because uf spatial aliasing, these procesrs ean perceive events with steep dips at high feequences as diferent oun what they actually are and, henee, do not trent hem prope ‘erly. For example, migration moves the spatially alias Ax=50m -10 to ieee (eyeles/kem) FIG. 1.20, A single, isolated dipping event and tts 2. ‘mpltde spectrum. Frequencies beyond 21 He ate spatially ‘ised. frequency components inthe wrong direction and gen ‘rates dsperive noise that degrades the quality of the migrated section. How is spatial aliasing avoided? Compare the see- tions in Figures 1.28 and 1.29, Both have the same {requency content, 6 to 42 Ha. The data in Figure 12-9, are spatially alinsed beeaise the dipping events toeper than in Figure 1.28. Some ways to avoid spatial alising follow: (@) Apply time shits so that the steep events appear to have lower dips. Although this oad change the lips that were low to higher dipe, making them spatially aliased, it often is & feasible solution forFundamentals of Signal Processing ar He 96 eo Ey 10 0 10 Wavenumber (cycles/km) FIG, 1.2-10. Six events with Keel dip nthe tx doe tual nip oot these radi line in the Jf domain. No frequencies are liner. Te dlpf thee even ithe arn o> ‘that ofthe nga even in Figure 12-8, What citings ‘his spectrum fom that in Figute 2-82 certain stations, For instance, «Tinear moveot ‘ea be applied to CMP gathers to eeumvent spa- ‘a aliasing of shallow events in prestack migration (Lier al, 1901). Ako, a hyperbolic moveout can be applied to COMP gathers before multiple atten ‘uaon inthe f —& domain (Setion 62), (8) Ta low-pass fer wore applied tothe traces in Fig- ‘ure 1.29 so that the frequencies up to 21 Ha were fetained, then the segment that is wrapped aronnd to the negative quacrant of the amplitude spec- ‘rum bs emowed. Although spatial aliasing is elim= ‘nated, a significant past of the recorded frequency Dand is lest. This approach isnot desirable. Wavenumber (cycles/km) PIG, 1.2-1. Six events with identical dip in the t— 2 do- ‘main map onto the sane ail ie the JK domain. Fee tunis beyond 21 Hove alias. The dip of these events 1S the same a that ofthe single event in Figure 12 {¢) Figure 12-18 shows a single dipping event reared with tees diferent trace spachugs. Tho 2-D amplie tude spectra sugarst a tied approach to solving ‘the spatial aliasing problem. Nove thatthe coarser the ace spacing, the more frequencies re spa- tially alise. The some frequency bandwidth is leptin oll three enves. The 12.5.0 trace spacing provides fequency band with no spetil aliasing. For a 25-1 trace spacing, frequencies beyond 36 Hz tre spatially alias: while for a 50-m trace spac- ing equencles beyond 18 He are spatially alias. or this later cate, spatial alinsing is 20 severe ‘that the aiased frequeneles wrap round the wave-Input Data, P(x) Plx0)= [ P(x) expCiaa)ae 4 10 Fourier Transform pac bree Mba P(e, plik, a)dx FIG. 1.212, Compunuon of the 2.0 Fora random, ‘mmber axis tie We see that spatial aliasing can be avoided by selocting a sulicienty stall ‘race spacing. This approach requires ether acata- Alependent interpolation selene (Section 7.2) to generate extra traces or modifiestion of the Feld recor geomet. Ifthe Inter approach ere taken, more shots and/or more rearing channels are needed To circumvent. spatial aliming, data often are recorded with twice as many number of ehauinels per shot record as tnt would be sed in processing. Prior tocdopping every other trace atthe start of a process ing seqence, a wavenuter filter is appa to remove the wavenumber components fr all Requcuces beyond "he Nyquist wavennmber that corsespends to the trace spacing of the data after dropping alternating traces from the shot rocord. This wavenuuber filter sctualy i a spatial high-eutantialine filter skis to the high-eut antilias frequency filter disersed in Setion 1.1 Consider the shot record shown in Figure 1.2-L4a recor! with 868 channels at n group iuterval of 1235 tm. The record which is intended fr input tov proces ing sequence comprises 192 channels with 2 sion Interval. Just as dropping every other sample i atime. series can result in frequency aliasing of sane of te high frequencies, cropping alterating tres cl result spatial aliasing (Pigure 12-14), Note that spatial alias. ing in the orginal record (Figure 1.2-L4a) occur start ing at about 75 hx. Trace decimation without spatial ‘anil filter causes spatial alnsing to eur stating at about 40 He (Figure 12-148), To circumvent spatial aliasing as a sesut of drop ping every ather trac inthe recor, a prior application of & spatial high-cutantialias filter is nese (Figure 1.2.10). Apply w wavenumber filter to remove the en: ergy within the J'— K region that iy between 50 and 100 wavenumber in both quadrants ofthe # — plane amie Data Analysis (Figure 12-154). Sine the trace spacing of the original feoord in Figure 1.2-14a is 12.5 ni, the corresponding Nyquist wavenumber is 40 cyeles/lan. Fallowing trace Secimation, the Nyquist wavenumber i 20 eyeles/km, cutresponding to 25-m trace spacing, Note in the J —k spectrum ia Figure 12-15 thatthe wanventimber filter has removed the energy betweets 20 and 40 eyeey/m forall frequencis in both quivennts, Compare the J —k spectra ofthe decimated data without (Fignte 12-14) and with antaos Altering (Figare 12-196) sud note ‘hat the wavenumber filter has removed te spatial alte high-roqueney components of the steeply dlp. Ping events inthe original gather. ‘The spatial antialas iter design aut application involves normi-inoveont coreetion of the data (Soe ‘on 8.1), Consider the 240-chamtel shot record shown in Figure 12-16a with 6.25-m group iterval We want te apply spatial antalas filter and reduce the numer of channels by half with 125m group inter. Pollen ‘the stop below: (a) Apply normal-moveost corection to the input xather (Figure 1.2-16a) that needs tobe decimate, to unaling the alinsed energy, (b) Apply 2-D Fourier tmnsform to map the data oto the fk plane. Follosing the normal-moweout cor roction, the energy maps in the f ~& plane cleser to the frequency axis ths alleviating the spatial aliasing (igre 12-17). (cl Design a spatial antialins filter by specifying re Jct ones for all feqencies between 50 and 100, \wavenmber in oth quadrants of the fk plane, (4) Apply the wavenumber filter to the moveout. forreted data in tho J —k domain (Figure 12- mm). (6) Apply inverse Foutiee transform of the waventimber-fitered data (Figure 12-108). (9, Now drop every other race to obtain the deviated data (Figure 12-16) The Nyquist yaweunmnbrer of ‘the gather fallowing the application ofthe spatial antialis ter sx oneal ob the Nyse maven ber of the original gather (Figure 2-17) (2) Apply inverse mvtont correction (Figure 1.2-164), ‘The testing shot recon! contains oue-hall the ‘umber of channel as inthe orginal ecard (igure 1.2-16n), and its energy is within the f — region defined by one-half the Nyquist wavenumber ass ciated with the original record (compare Figures 12-470 and 12-174 The original moveout-corrected shot record (Fig- ue 12-10) alter dropping every other trace, but with ‘out the application of the antinlins waserminber filter, is horn in Figure 12160 with the J # spectrum ofFundamentals of Signal Processing Trace Spacing 12.5m 25m -40 0 40-20 0 20-10 0 10 Wavenumber (cycles/km) FIG. 12-18, A single, Solotd dipping event sample at thre diferent scce spacings with the corresponding f— spectra ‘No spatial aliasing occurs with the 12. trace spacing (ft). Preueneis beyond 8 Hz ao alse with 1 pacng (eater), Double aliasing occurs with the Sl-m trae seeing (ight). Although events on the f ~K spectra Juno diferent dips, thre have the sae dip an he ¢ tur (top). This deceptive character Is because of the erent oriaoitalsenles used in displaying to f ~# spectra 5960 Data Analysis won _a7275m Fa pHi ; ae aa @ ° FIG. 1.2-14, (a) shot record with ts 24D amplitude spactram esplaved on tops trace sping is 12.5 m and the Nyquist ‘wnvenumbee i 40 eyten/ken. Minimum aoa maimum ots age 1am and 47275 mn eapetively,(b) The same exon 8 in (a) with every other trace dropped; henew, trace spacig i 25 m and the Nyquist wavenumber 20 eycles/kam, Note that tte enezgy is spatially lied above 75 Tl in the ceginal record (a), while spatial aliasing occurs beginning a1 Hein the cord with every othe trast dropped (6)Fundamentals of Signal Processing 6 eesti yseten ® FIG. 1.2.15. (a) The sane shot record ati Figaro 12-140 afcr che appieaton of wavetumbor Mle o remove the energy Devon half the Nyqube wavener forall frequencies (0) The sane record a in (a) alter dropping every other trace. Nove ‘hat spatial aang i cireurvented et by appiving a high-cut spatial atilasing wavemumber Site (a) pr to dropping trey ote taco, In conte, rpg ery other race without the peor application of an antialos ler causes spt Alias of events with large move (Figure 12-13)Seismic Data Analysis 62Fundamentals of Signal Processing — = ais , : = ic e Id f 40 0 40 FIG. 1.2-17, The f — spect ofthe shot record at vations snes as deseribod in Figure 1.216oa Seismle Data Analysis PIG. 1.218. A serootset ston (256 trace with 250 trace spacing) containing 1D dipping events and ie 2D ae plitude spectrum. No euencis ate alse. ‘the resulting record shown in Figure 1.2-17e. This dei- ‘mated shot record (Figure 12-16e) should be compare with the antialiveivered record (Figure 12 16e)af- {er inverse moveout eorectlan (Figures 12-16 and {) Examine the f—& spectra ofthe datn afer trace deci sation with (Figure 12-174) and without antalies fie tering (Figure 12-171) and note that, for this data set, ‘wavenumber filtering has not been succesful in eno." ing spatially aliaseel energy. Waveniimber Silzering for trace decimation is successful if the aliased energy in {he original record js confined to the eset ones fr all frequencies betwen 50 ad 100, So far, only the ynthess ofa single dipping event from a déereve number of frequeney components has ‘een considered. This analysis now i extended t0 a range of dips. Figure 12-18 shows a section with dips that vary from 0 to 45 degrees andl the eoresponding 2-D amplitude spectrum. These sae dips, Int with Iigher trequeney content, also are sen in Figure 12- 19. Events with 0+, 5, 10, and 15-degroe dips ae not spatially aliased. Tho 20-degree dip i aliased at nearly 72 Ha, the B0cdegre dip at nearly 48 He, and the 45- degre dip at nearly 36 He. Again, the steeper the dip, ‘he lower the frequency at which spatial aliasing oceur eet) m Waverumber (eyeles im) FIG. 1.2-19. A zerother section (256 tres with 26m trace spacing) containing 10 dipping events ads 2D a plitude spectrum, Scecper dip ate lied © nereasingly Tower fequeacie Given a dip valuc, how isthe maxinmutn maliased frequency determined? Consider the 20-degree dipning ‘event in Figure 1.2.19. First, measute the dip i ails fonds per trace. There are 256 traces in the = model with 25m trace spacing. The 20-degree dip is equivalent ‘to T ms/trace, Frequency components with periods less ‘than twice the dip are spatially alias. Thus, siven the ip in milieonds per trace, the Useeold frequency ‘at which spatial allesing begin is 500 per dip. a the present case, che threshold feequeney is 00/7 = 72 He, ‘This is verified by examining the amplitude spectrum in Figure 12-19. Figure 12-20 shows three field records sd thei 2 D amplitude spectra, nown as f — spectra. By now, it is easy to recogniae and relate various events on the shot gathers to those on the f — & spectra. Event As the high-amplitude dispersive coberent noise with very low group wlocity. When the spatial extent of these ‘waves broadens, bandwidth in the savensmber ditee- tion becomes smaller. Conversely, whet the spatial ex teat becomes sanaller, the event, such as G, spans a wider yovenumber bandwidth in the f — & spectrum (compare events A, F, and G), Events Brand C are partsFundamentals of Signal Processing 65 FSR SSIS FIG. 12-20. Thee common-hot gashers (top) an their f— spect 36 24 12 lz (bottom). (Ve mark events ate discs in the text) Dip cnveatin: Au event maps onto a pesir-dip quadrant inthe fk spectra if dips down moving fom near {far offs (Data courtesy Deine Petrleu.) of the guided wave packet. Event © contains alse energy above 42 He (indicated by D on the J — k spec: ‘um), Primaries avd associated multiples are mapped Into region Ebetwoen the frequeney axis and event C. Spatial aliasing not only isa concer in a prestack Application of a mmitichanael flker, each as f — & fle tering, but also during: poststack processing. Specif- ‘aly, migration of CMP stacked data suffers fren spa tial aliasing ifthe trace interval is too conrae. An optic ‘mim CMP trace interval ta avoid spatial aliasing ea be computed a follows. Consier a dipping telector ‘vith a dip angle of @ (Figure 1.221). Also comsider normalnidenee plane wave with» dominant period TT recorded at the surface with a trace sepa (This s the ero-ffut ease whore A is the CMP trace Interval) From the geometry in Figure 1.221, we write sind a-6 ‘At ae ‘whore AY Is the twos time separation between the antival mes of the plane wave at the two recciver Io- tations, A and B, ane» isthe mean velocity. Spatial66 Seismic Data Analysis FIG. 1.2:21 A plane wave reflecting st sormal aeldenc om reocver lations and Bat the sirace with ‘lzsing occurs when the wavefront separation in tine AL equals half the dominant period 7. When this er terion is applied to equation (I-), we get the follow iow expression for the maximum threshold frequency, Sr: = YT, that 18 not aliased for a given dip. weloe, ityeand CMP trace interval: Sawn = a-7 ‘Table 1-12 shows the evaluation of equation (1-7) for particular mamerical example, Equasion (1-7) aso can be expressed in terms of seeciver group interval 24x. Suppose the maximum dip is 30 deg. If the sampling interval ls 4 ms, then the Nyquist frequency bs 125 Hz. After antalas filtering, the trequeney bund fextonds up to 90 Hz, provided the high-eut filter i at ‘hres-quarters of the Nyquist. Fora bandwidth without spatial aliasing, we have to select 4 12.5:m CMP trace interval ‘To circumvent the deleterions effect of spatial alias Ing on migrated data, trace interpolation i applied to attain a trace intereal appropriate for the bandwidth ofthe CMPestacked data. As tated in Section Lt, ED interpolation ofa time series, sich as a sisnie trace from a coarse to a finer sampling iterval docs not re. Aippng elector with dip ange 0. partion Az. Geometry ofthis plane wave i ued to drive oquation (1), ‘Table 1-12. Frequency threshold for spatial aliasing Velocity is 3000 m/s ‘Threshold Froqueney (Ha) for CMP Trace Tnterval (tn) Dip Angle (eg) 12525 750 0 Me In us 86 20 1% 8 a 0 60030 40 ee) ‘cover the frequents Jost by the original sampling: but, fs only generates extn sumples. On the other Ind, ine texpolation of 2-D data, such as «seismic section, from 4 coarse to a finer trace interval, is posible forall sige tal frequencies, ncding those that ar alised This is because ofthe fat that one ean detect and measure the dip as defined by the stepout of equation (1-5) on stacked section fr the unaliase fequencies and se the ‘ip information to interpolate not only the unaliased bt also the aliased frequencies. Modern data acquis tom geometries for 2-D data often do not renie teaFundamentals of Signal Processing or Interpolation. Nevertheless, trace interpolation sally ‘seme inthe erent direction prcr to -D migra: tion (Section 7.2) 1.3 WORLDWIDE ASSORTMENT OF SHOT RECORDS For shot recotds, both land nd marae, om Noch fund Soh Ameria, Enrope, the Mile Ent. North ‘Alten, and the Far East re presented in Figs 1 1 throwgh 1-40. Sonre types are viboxcln, Ceol. snauite sit gin, Maspnne, Aqeapue, and Aqua: fel, The reading parameter, Ineluding the ruber Of tuces, umber of samples per tac, samp i feral ace interval diner of. ae nda “able 11. Stay the Sek ecto haa how to ee ize diferent types of waves. For dp pres thinstantaneous type of gain (AGC) (Section 1A) wat Applied to all 40 coeds. These recor wll be rete toby thelr read numer in the talon actions "he nnn oa in procesing ellen sec data 1s to enhance gemsine reflection signal hy mppmesing uy energy he orm of coherent sa rand bint noise Inthe following paragraphs, sot gathers fo examined Yo pint out the diferent types of seme erg. Thecord 1 a corel vitro dat st. (For ives eorelatio, veer to Section 1.) A number Of seftions i presatinthis rear vith reasonably fc igaltonsise ratio. A genine reflection see redo coniuonabot gathers by its iypebotle nate Reflotions belive the sae way on CMP gather. A fat horizon with wy dip sielde symmetichyperiola oa both conmon-het aid emmowsmidpint gothers ecorded using splitspread prot (a aplt-speend zomety, the sotcei oeted somewhere fn the mie Ale of the receiver ene, usualy a the center) p= ing hora yes a skewed ype on a co tok gather, white stil silling 8 symmetic hyperbola 0 a commoirmidpint gather. Repro of soe fd testers provides this symmetry. From the r= fon hyperbas in Record, note tt he subeeface ‘nama up of weanly horizontally fa layers. Any ee iit inthe shape of the moveoutlyperbola can be tributed to na-srface ect adr Intra ara tions in veo Record 2 is an asymmetee shot gather. Note the ‘elation energy betwen 1nd 2 with athe ea lar monet. Record 8, which was obtained by using dynanie, Contains a series of rections with nealy ere Datloic moveoit, especially between 1 ahd 3 sis record is from the times of analog recording, I x not ‘meommoe to digitize old snag dats and procs ie ‘with modern teetniqaes, ‘Record 4 contains events with complex maveout between 25 aul 3.5 s Events A,B, and C have skewed hyperbolic moweauts, whieh, n Ths ease, stguests that ‘ey are dipping up torard the left Also note to travel time distortions along the moveouts caused, most My, by near-surface iregularitie. Recor § has some grounderoll energy, whieh is iacactrizod by its low-frequency, high-amplitude ap- pearance, particularly on short-offer traces This kind of energy typically is suppressed in the fel by using a proper reeiver array ‘Record 6 contains wea snd strong, neatly at re fectors (A and B). The break in the reflection hyper- bola (C1-02 } suggests the presence of a fault C1 on the upthrown side and C2 on the downthrown side) Again, note the ground-roll energy with ie dispersed low-frequency character on inside traces (event D). econ 7 contains threo interesting events. Event A ia skewed hyperbola, which suggests that it dipping up toward the left, while event B is nearly syinmetee, Which suguests lat dip. Event C shows a discontinity F along ite mosvont curve, indenting the presence of faa Reconl 8 shows a ror with exellent signal qual ity, This dynamite record bas « mumbor of reiletions land associted infeed reverberations. Note the pro- aresive deereave in the sigabto-noive ratio at late times. This ie true for alment all sexe data. Event ‘Alias largo moveout because i is shallow, while event B has amall moveout beenuse iti dep, (Linear energy Cis referred to in Exercise 1-3.) Tecord 9, which js « eoreiated vibroses dataset, Jn a sris of rfloctions and ground roll. Unlike daca fom impulsive sources such as dynamite, fist breaks in vibrosis data may not be distinguishable (compare, for example, Reeotds 8 and 9). This s because the cor related vibrosels record contains some ofthe side lobes Of the sweap signal autocorrelation. Note the increase Jn random noise in the later part ef the record helow 3s Rocord 10 contains two strong shallow releesoes, ‘A and B, in addition to gromel-oll energy C. Alo, & bundle of energy with extremely large moveout is noted between 25 and 5s (DI-D2). This coherent noise may be attributed to sidescattered energy, whlch & eased by inhomogenctis in the subsurface (particularly at the water bottom) thet behave as point sourees. Recor 11 contains four prominent reflections. This record Is from Alaska, where the thickness of the per maftost layer ean be iregula. Such near-surface ize6s ‘Table 1-18, Parameter Index ofa worldwide aswortment of comms Recont Naner 2SeuNeueen e4eeeee 30 0 Area South Texas West Texas Lowi ‘Turkey South America Far Bast South Amerien (Central Ameren Alas North Aiea Alaska Missin Ofshoce Texas Offshore Texas Offshore Canada South Ameren South Amerien Oiitiore Louisiana Turkey Sonth Aleutians Denver Basin Williston Basin San Jaci Basia Arctic Alberta Alberta Canada Canada Olkhore Span Offshore Crete North Sea North Sea North Sea Celie Sen Denmark Midale East Turkey North Aftea Mile Fast West Aiea Seismic Data Analysis g ot gathers, ‘Number of Trace aver ‘Saumpke Number of Sampling Taterval, Off, er Trace ‘Traces Interval, ms eorm form 1,275 48 4 son ovo 1035 120 4 Joo 400 4300 a 4 Mom 310% 175 8 4 100m 350m 3.000 a 2 oom 300m 1.350 4 4 100m 150m 2.600 8 2 100m 360m 1300 06 4 som 100m 10 % 4 0k sao 1335 120 ‘ 2m 30m 1000 96 4 2% oon 1275 8 . BOR oon 20% As ‘ DR TN 1535 4s ‘ 20 ooo 2,500 8 2 25m asihin 1305 4 4 2m 283m 2.000 4s 4 50m 250m 1,50 120 ‘ eA To 4,250 216 4 rt} 50m 20% 120 4 mR oh 1,550 18 2 zor m0 1550 a8 2 ok 0 1380 8 2 Don on 8.000) is 2 200 Bon 2,000 % 2 om 50m 1.500 8 2 om orm 1,701 » 4 Som 200m 2.500 as 2 2m 00m 2.000 4s 4 50m 250m 2125 9% 4 2m 11360 6 fi 25m 1550 % 4 25m 1625 % 4 25m 1500 « 4 50m 2.500 3 2 100 2 1024 & 4 om 1.000 4 4 Tm 250 ry 2 100 m 2.500 oo 2 50m 2.600 % 2 30m COCK UD ppb peep co uccchuseee we

You might also like

• The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  

  From Everand

  The Subtle Art of Not Giving a F*ck: A Counterintuitive Approach to Living a Good Life

  Mark Manson

  4/5 (6058)
• Principles: Life and Work

  

  From Everand

  Principles: Life and Work

  Ray Dalio

  4/5 (626)
• The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  

  From Everand

  The Gifts of Imperfection: Let Go of Who You Think You're Supposed to Be and Embrace Who You Are

  Brene Brown

  4/5 (1142)
• Never Split the Difference: Negotiating As If Your Life Depended On It

  

  From Everand

  Never Split the Difference: Negotiating As If Your Life Depended On It

  Chris Voss

  4.5/5 (918)
• The Glass Castle: A Memoir

  

  From Everand

  The Glass Castle: A Memoir

  Jeannette Walls

  4.5/5 (1752)
• Sing, Unburied, Sing: A Novel

  

  From Everand

  Sing, Unburied, Sing: A Novel

  Jesmyn Ward

  4/5 (1245)
• Grit: The Power of Passion and Perseverance

  

  From Everand

  Grit: The Power of Passion and Perseverance

  Angela Duckworth

  4/5 (628)
• Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  

  From Everand

  Hidden Figures: The American Dream and the Untold Story of the Black Women Mathematicians Who Helped Win the Space Race

  Margot Lee Shetterly

  4/5 (946)
• Shoe Dog: A Memoir by the Creator of Nike

  

  From Everand

  Shoe Dog: A Memoir by the Creator of Nike

  Phil Knight

  4.5/5 (550)
• The Perks of Being a Wallflower

  

  From Everand

  The Perks of Being a Wallflower

  Stephen Chbosky

  4.5/5 (2124)
• The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  

  From Everand

  The Hard Thing About Hard Things: Building a Business When There Are No Easy Answers

  Ben Horowitz

  4.5/5 (361)
• Her Body and Other Parties: Stories

  

  From Everand

  Her Body and Other Parties: Stories

  Carmen Maria Machado

  4/5 (831)
• Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  

  From Everand

  Elon Musk: Tesla, SpaceX, and the Quest for a Fantastic Future

  Ashlee Vance

  4.5/5 (482)
• Steve Jobs

  

  From Everand

  Steve Jobs

  Walter Isaacson

  4.5/5 (822)
• The Emperor of All Maladies: A Biography of Cancer

  

  From Everand

  The Emperor of All Maladies: A Biography of Cancer

  Siddhartha Mukherjee

  4.5/5 (277)
• Bad Feminist: Essays

  

  From Everand

  Bad Feminist: Essays

  Roxane Gay

  4/5 (1066)
• The Little Book of Hygge: Danish Secrets to Happy Living

  

  From Everand

  The Little Book of Hygge: Danish Secrets to Happy Living

  Meik Wiking

  3.5/5 (438)
• Angela's Ashes: A Memoir

  

  From Everand

  Angela's Ashes: A Memoir

  Frank McCourt

  4.5/5 (444)
• Brooklyn: A Novel

  

  From Everand

  Brooklyn: A Novel

  Colm Toibin

  3.5/5 (2037)
• A Man Called Ove: A Novel

  

  From Everand

  A Man Called Ove: A Novel

  Fredrik Backman

  4.5/5 (4971)
• The Yellow House: A Memoir (2019 National Book Award Winner)

  

  From Everand

  The Yellow House: A Memoir (2019 National Book Award Winner)

  Sarah M. Broom

  4/5 (99)
• Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  

  From Everand

  Devil in the Grove: Thurgood Marshall, the Groveland Boys, and the Dawn of a New America

  Gilbert King

  4.5/5 (273)
• The Art of Racing in the Rain: A Novel

  

  From Everand

  The Art of Racing in the Rain: A Novel

  Garth Stein

  4/5 (4273)
• Yes Please

  

  From Everand

  Yes Please

  Amy Poehler

  4/5 (1971)
• The World Is Flat 3.0: A Brief History of the Twenty-first Century

  

  From Everand

  The World Is Flat 3.0: A Brief History of the Twenty-first Century

  Thomas L. Friedman

  3.5/5 (2283)
• The Outsider: A Novel

  

  From Everand

  The Outsider: A Novel

  Stephen King

  4/5 (1957)
• The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  

  From Everand

  The Sympathizer: A Novel (Pulitzer Prize for Fiction)

  Viet Thanh Nguyen

  4.5/5 (125)
• A Tree Grows in Brooklyn

  

  From Everand

  A Tree Grows in Brooklyn

  Betty Smith

  4.5/5 (1934)
• Team of Rivals: The Political Genius of Abraham Lincoln

  

  From Everand

  Team of Rivals: The Political Genius of Abraham Lincoln

  Doris Kearns Goodwin

  4.5/5 (236)
• The Woman in Cabin 10

  

  From Everand

  The Woman in Cabin 10

  Ruth Ware

  3.5/5 (2619)
• A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  

  From Everand

  A Heartbreaking Work Of Staggering Genius: A Memoir Based on a True Story

  Dave Eggers

  3.5/5 (233)
• Wolf Hall: A Novel

  

  From Everand

  Wolf Hall: A Novel

  Hilary Mantel

  4/5 (4060)
• On Fire: The (Burning) Case for a Green New Deal

  

  From Everand

  On Fire: The (Burning) Case for a Green New Deal

  Naomi Klein

  4/5 (75)
• Rise of ISIS: A Threat We Can't Ignore

  

  From Everand

  Rise of ISIS: A Threat We Can't Ignore

  Jay Sekulow

  3.5/5 (142)
• Fear: Trump in the White House

  

  From Everand

  Fear: Trump in the White House

  Bob Woodward

  3.5/5 (806)
• Manhattan Beach: A Novel

  

  From Everand

  Manhattan Beach: A Novel

  Jennifer Egan

  3.5/5 (885)
• Gold Leaf Manufacturing Process

  

  No ratings yet

  Gold Leaf Manufacturing Process

  5 pages
• The Unwinding: An Inner History of the New America

  

  From Everand

  The Unwinding: An Inner History of the New America

  George Packer

  4/5 (45)
• John Adams

  

  From Everand

  John Adams

  David McCullough

  4.5/5 (2520)
• The Light Between Oceans: A Novel

  

  From Everand

  The Light Between Oceans: A Novel

  M L Stedman

  4.5/5 (790)
• The Constant Gardener: A Novel

  

  From Everand

  The Constant Gardener: A Novel

  John le Carré

  3.5/5 (109)
• LW-781 (Cg08) DRL PGM One Page Rev

  

  No ratings yet

  LW-781 (Cg08) DRL PGM One Page Rev

  2 pages
• 2013-01-01-R-543 (NP - GE31) - Rumaila DOR

  

  No ratings yet

  2013-01-01-R-543 (NP - GE31) - Rumaila DOR

  2 pages
• Acoustic Waves and Tools: Baker Atlas Montrose Training Centre

  

  No ratings yet

  Acoustic Waves and Tools: Baker Atlas Montrose Training Centre

  34 pages
• Petroleum Systems

  

  No ratings yet

  Petroleum Systems

  20 pages
• 4 Silica Sand PDF

  

  No ratings yet

  4 Silica Sand PDF

  1 page
• Pnacs707 2

  

  No ratings yet

  Pnacs707 2

  173 pages
• 2111 PDF

  

  No ratings yet

  2111 PDF

  222 pages
• Silica Sand Manufacturers in Rajasthan

  

  No ratings yet

  Silica Sand Manufacturers in Rajasthan

  1 page
• Pnacs707 2

  

  No ratings yet

  Pnacs707 2

  173 pages
• The Petrology of The Abu Zawal Gabbroic Intrusion, Eastern Desert, Egypt: An Example of An Island-Arc Setting

  

  No ratings yet

  The Petrology of The Abu Zawal Gabbroic Intrusion, Eastern Desert, Egypt: An Example of An Island-Arc Setting

  11 pages
• Gemston Es: in Egypt

  

  No ratings yet

  Gemston Es: in Egypt

  1 page
• Well Data

  

  No ratings yet

  Well Data

  5 pages
• Insight Into African Meteorite Finds: Typology, Mass Distribution and Weathering Process

  

  No ratings yet

  Insight Into African Meteorite Finds: Typology, Mass Distribution and Weathering Process

  14 pages
• 03 The Study of Economic Importance of Granite Rock

  

  No ratings yet

  03 The Study of Economic Importance of Granite Rock

  33 pages
• AGMCO Prequalifications 2019

  

  No ratings yet

  AGMCO Prequalifications 2019

  21 pages
• 497 010 PDF

  

  No ratings yet

  497 010 PDF

  17 pages
• SPWLA 2002 Decoster CerroNegro

  

  No ratings yet

  SPWLA 2002 Decoster CerroNegro

  14 pages
• Little Women

  

  From Everand

  Little Women

  Louisa May Alcott

  4/5 (105)