Note no SAND/05/2012

Authors
Pl Dahle
Bjrn Fjellvoll
Frode Georgsen
Ragnar Hauge
Odd Kolbjrnsen
Anne Randi Syversveen
Marit Ulvmoen
Date March 2, 2012
CRAVA User Manual
version 1.2
Pl DahleBjrn Fjellvoll Frode GeorgsenRagnar HaugeOdd KolbjrnsenAnne Randi SyversveenMarit Ulvmoen
The authors
Pl Dahle is a Senior Research Scientist at NR,
Bjrn Fjellvoll is a Research Scientist at NR,
Frode Georgsen is a Senior Research Scientist at NR.
Ragnar Hauge is an Assistant Research Director at NR,
Odd Kolbjrnsen is a Chief Research Scientist at NR,
Anne Randi Syversveen is a Senior Research Scientist at NR and
Marit Ulvmoen is a Research Scientist at NR.
Norwegian Computing Center
Norsk Regnesentral (Norwegian Computing Center, NR) is a private, independent, non-prot
foundation established in 1952. NR carries out contract research and development projects in
information and communication technology and applied statistical-mathematical modelling. The
clients include a broad range of industrial, commercial and public service organisations in the
national as well as the international market. Our scientic and technical capabilities are further
developed in co-operation with The Research Council of Norway and key customers. The results
of our projects may take the form of reports, software, prototypes, and short courses. A proof of
the condence and appreciation our clients have in us is given by the fact that most of our new
contracts are signed with previous customers.
Title CRAVA User Manual version 1.2
Authors Pl Dahle, Bjrn Fjellvoll , Frode Georgsen, Ragnar Hauge,
Odd Kolbjrnsen, Anne Randi Syversveen, Marit Ulvmoen
Date March 2, 2012
Publication number SAND/05/2012
Abstract
CRAVA is a simple inversion tool, particularly suited for quick generation of rst pass inversions
and facies probabilities for use in geological modeling. This manual describes the theory behind,
the main implementation structure, and the actual use of this program.
Keywords CRAVA, seismic, inversion, geostatistical, Bayesian, AVO, FFT
Target group NR, Statoil, Norsar, Roxar
Availability Open
Project -
Project number -
Research eld Reservoir characterisation
Number of pages 93
Copyright 2012 Norwegian Computing Center
3
Contents
1 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 AVO . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.3 Seismic model . . . . . . . . . . . . . . . . . . . . . . . . 15
1.3.1 Convolution with 3D wavelet . . . . . . . . . . . . . . . 16
1.4 Statistical model . . . . . . . . . . . . . . . . . . . . . . . 16
1.4.1 Facies probabilities . . . . . . . . . . . . . . . . . . . 18
2 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.1 Estimating optimal well location. . . . . . . . . . . . . . . . . . 22
2.2 Estimating the prior model . . . . . . . . . . . . . . . . . . . 22
2.2.1 Background model . . . . . . . . . . . . . . . . . . . 22
2.2.2 Covariance. . . . . . . . . . . . . . . . . . . . . . 25
2.2.3 Likelihood model. . . . . . . . . . . . . . . . . . . . 25
2.3 Estimating wavelets . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Estimating 3D wavelet . . . . . . . . . . . . . . . . . . . . . 28
2.5 Using FFT for inversion. . . . . . . . . . . . . . . . . . . . . 28
2.6 A note one local wavelet and noise . . . . . . . . . . . . . . . . 28
2.6.1 Local wavelet - dividing out the wavelet . . . . . . . . . . . 28
2.6.2 Local noise. . . . . . . . . . . . . . . . . . . . . . 29
2.7 Memory handling . . . . . . . . . . . . . . . . . . . . . . . 29
2.7.1 Grid allocation with all grids in memory . . . . . . . . . . . 29
3 User guide . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.1 Basic inversion. . . . . . . . . . . . . . . . . . . . . . . . 31
3.1.1 Survey information . . . . . . . . . . . . . . . . . . . 31
3.1.1.1 Seismic data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.1.2 Wavelet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.1.3 Signal/noise ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.2 Inversion volume. . . . . . . . . . . . . . . . . . . . 33
3.1.2.1 Lateral extent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.1.2.2 Top and base surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.1.2.3 Depth conversion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
CRAVA User Manual version 1.2 4
3.1.3 Prior model . . . . . . . . . . . . . . . . . . . . . 36
3.1.3.1 Background model . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1.3.2 Covariances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.4 Well data . . . . . . . . . . . . . . . . . . . . . . 37
3.1.5 I/O settings. . . . . . . . . . . . . . . . . . . . . . 38
3.1.6 Output . . . . . . . . . . . . . . . . . . . . . . . 38
3.1.6.1 Grid output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.6.2 Well output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.1.7 Actions . . . . . . . . . . . . . . . . . . . . . . . 40
3.1.8 Standard grid formats . . . . . . . . . . . . . . . . . . 40
3.2 Advanced inversion options . . . . . . . . . . . . . . . . . . . 41
3.2.1 Non-stationary wavelet and noise . . . . . . . . . . . . . 41
3.2.2 PS-seismic and reection approximations. . . . . . . . . . . 41
3.2.3 Well quality checks . . . . . . . . . . . . . . . . . . . 42
3.2.4 Generate synthetic seismic from inversion data . . . . . . . . 42
3.3 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Estimation mode. . . . . . . . . . . . . . . . . . . . 42
3.3.2 Wavelet and noise estimation . . . . . . . . . . . . . . . 43
3.3.3 Background model estimation . . . . . . . . . . . . . . . 43
3.3.4 Prior correlations . . . . . . . . . . . . . . . . . . . 44
3.4 3D wavelet . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5 Facies prediction . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.1 Prior probabilities . . . . . . . . . . . . . . . . . . . 46
3.5.2 Output parameters . . . . . . . . . . . . . . . . . . . 46
3.6 Forward modelling . . . . . . . . . . . . . . . . . . . . . . 46
4 Model le reference manual . . . . . . . . . . . . . . . . . . . . . 48
4.1 <actions> (necessary) . . . . . . . . . . . . . . . . . . . . . 48
4.1.1 <mode> (necessary) . . . . . . . . . . . . . . . . . . . 48
4.1.2 <inversion-settings> . . . . . . . . . . . . . . . . . 48
4.1.2.1 <prediction> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.1.2.2 <simulation> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.1.2.2.1 <seed> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.1.2.2.2 <seed-file> . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1.2.2.3 <number-of-simulations> . . . . . . . . . . . . . . . . . 49
4.1.2.3 <kriging-to-wells> . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1.2.4 <facies-probabilities> . . . . . . . . . . . . . . . . . . . . . . . 49
CRAVA User Manual version 1.2 5
4.1.3 <estimation-settings>. . . . . . . . . . . . . . . . . 49
4.1.3.1 <estimate-background> . . . . . . . . . . . . . . . . . . . . . . . . 49
4.1.3.2 <estimate-correlations> . . . . . . . . . . . . . . . . . . . . . . . 49
4.1.3.3 <estimate-wavelet-or-noise> . . . . . . . . . . . . . . . . . . . . 49
4.2 <project-settings> (necessary) . . . . . . . . . . . . . . . . . 49
4.2.1 <output-volume> (necessary) . . . . . . . . . . . . . . . 50
4.2.1.1 <interval-two-surfaces> . . . . . . . . . . . . . . . . . . . . . . . 50
4.2.1.1.1 <top-surface> (necessary) . . . . . . . . . . . . . . . . . . 50
4.2.1.1.1.1 <time-file> . . . . . . . . . . . . . . . . . . . . . . 50
4.2.1.1.1.2 <time-value> . . . . . . . . . . . . . . . . . . . . . 50
4.2.1.1.1.3 <depth-file> . . . . . . . . . . . . . . . . . . . . . 50
4.2.1.1.2 <base-surface> (necessary) . . . . . . . . . . . . . . . . . 50
4.2.1.1.2.1 <time-file> . . . . . . . . . . . . . . . . . . . . . . 50
4.2.1.1.2.2 <time-value> . . . . . . . . . . . . . . . . . . . . . 51
4.2.1.1.2.3 <depth-file> . . . . . . . . . . . . . . . . . . . . . 51
4.2.1.1.3 <number-of-layers> . . . . . . . . . . . . . . . . . . . . . 51
4.2.1.1.4 <velocity-field> . . . . . . . . . . . . . . . . . . . . . . 51
4.2.1.1.5 <velocity-field-from-inversion> . . . . . . . . . . . . 51
4.2.1.2 <interval-one-surface> . . . . . . . . . . . . . . . . . . . . . . . 51
4.2.1.2.1 <reference-surface> . . . . . . . . . . . . . . . . . . . . 51
4.2.1.2.2 <shift-to-interval-top> . . . . . . . . . . . . . . . . . 51
4.2.1.2.3 <thickness> . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.1.2.4 <sample-density> . . . . . . . . . . . . . . . . . . . . . . 52
4.2.1.3 <area-from-surface> . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.1.3.1 <file-name> . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.1.3.2 <snap-to-seismic-data> . . . . . . . . . . . . . . . . . . 52
4.2.1.4 <utm-coordinates> . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.1.4.1 <reference-point-x> . . . . . . . . . . . . . . . . . . . . 52
4.2.1.4.2 <reference-point-y> . . . . . . . . . . . . . . . . . . . . 53
4.2.1.4.3 <length-x> . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.1.4.4 <length-y> . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.1.4.5 <sample-density-x> . . . . . . . . . . . . . . . . . . . . . 53
4.2.1.4.6 <sample-density-y> . . . . . . . . . . . . . . . . . . . . . 53
4.2.1.4.7 <angle> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.1.4.8 <snap-to-seismic-data> . . . . . . . . . . . . . . . . . . 53
4.2.1.5 <inline-crossline-numbers> . . . . . . . . . . . . . . . . . . . . . 53
4.2.1.5.1 <il-start> . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.1.5.2 <il-end> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
CRAVA User Manual version 1.2 6
4.2.1.5.3 <xl-start> . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.1.5.4 <xl-end> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.1.5.5 <il-step> . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.1.5.6 <xl-step> . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.2 <time-to-depth-mapping-for-3d-wavelet> . . . . . . . . . 54
4.2.2.1 <reference-depth> . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.2.2 <average-velocity> . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.2.2.3 <reference-time-surface> . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3 <io-settings> . . . . . . . . . . . . . . . . . . . . 55
4.2.3.1 <top-directory> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.2 <input-directory> . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.3 <output-directory> . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.4 <grid-output> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.4.1 <domain> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.4.1.1 <depth> . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.4.1.2 <time> . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3.4.2 <format> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.2.3.4.2.1 <segy-format> . . . . . . . . . . . . . . . . . . . . 56
4.2.3.4.2.1.1 <standard-format> . . . . . . . . . . . . . . 56
4.2.3.4.2.1.2 <location-x> . . . . . . . . . . . . . . . . . 56
4.2.3.4.2.1.3 <location-y> . . . . . . . . . . . . . . . . . 56
4.2.3.4.2.1.4 <location-il> . . . . . . . . . . . . . . . . 56
4.2.3.4.2.1.5 <location-xl> . . . . . . . . . . . . . . . . 56
4.2.3.4.2.1.6 <bypass-coordinate-scaling> . . . . . . . 56
4.2.3.4.2.1.7 <location-scaling-coefficient> . . . . . 57
4.2.3.4.2.2 <segy> . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.2.3 <storm> . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.2.4 <crava> . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.2.5 <sgri> . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.2.6 <ascii> . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.3 <elastic-parameters> . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.3.1 <vp> . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.3.2 <vs> . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2.3.4.3.3 <density> . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.4 <lame-lambda> . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.5 <lame-mu> . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.6 <poisson-ratio> . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.7 <ai> . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
CRAVA User Manual version 1.2 7
4.2.3.4.3.8 <si> . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.9 <vp-vs-ratio> . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.10 <murho> . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.11 <lambdarho> . . . . . . . . . . . . . . . . . . . . . . 58
4.2.3.4.3.12 <background> . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.4.3.13 <background-trend> . . . . . . . . . . . . . . . . . 59
4.2.3.4.4 <seismic-data> . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.4.4.1 <original> . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.4.4.2 <synthetic> . . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.4.4.3 <residuals> . . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.4.4.4 <synthetic-residuals> . . . . . . . . . . . . . . . 59
4.2.3.4.5 <other-parameters> . . . . . . . . . . . . . . . . . . . . . 59
4.2.3.4.5.1 <facies-probabilities> . . . . . . . . . . . . . . 59
4.2.3.4.5.2 <facies-probabilities-with-undef> . . . . . . 60
4.2.3.4.5.3 <facies-likelihood> . . . . . . . . . . . . . . . . 60
4.2.3.4.5.4 <time-to-depth-velocity> . . . . . . . . . . . . . 60
4.2.3.4.5.5 <extra-grids> . . . . . . . . . . . . . . . . . . . . 60
4.2.3.4.5.6 <correlations> . . . . . . . . . . . . . . . . . . . . 60
4.2.3.4.5.7 <seismic-quality-grid> . . . . . . . . . . . . . . 60
4.2.3.5 <well-output> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.3.5.1 <format> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.5.1.1 <rms> . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.5.1.2 <norsar> . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.5.2 <wells> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.5.3 <blocked-wells> . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.5.4 <blocked-logs> . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.6 <wavelet-output> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.6.1 <format> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.6.1.1 <jason> . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.3.6.1.2 <norsar> . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3.6.2 <well-wavelets> . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3.6.3 <global-wavelets> . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3.6.4 <local-wavelets> . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3.7 <other-output> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3.7.1 <extra-surfaces> . . . . . . . . . . . . . . . . . . . . . . 62
4.2.3.7.2 <prior-correlations> . . . . . . . . . . . . . . . . . . . . 62
4.2.3.7.3 <background-trend-1d> . . . . . . . . . . . . . . . . . . . 62
4.2.3.7.4 <local-noise> . . . . . . . . . . . . . . . . . . . . . . . . 63
CRAVA User Manual version 1.2 8
4.2.3.7.5 <rock-physics-distributions> . . . . . . . . . . . . . . 63
4.2.3.7.6 <error-file> . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.3.7.7 <task-file> . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.3.8 <file-output-prefix> . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.3.9 <log-level> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.4 <advanced-settings> . . . . . . . . . . . . . . . . . 63
4.2.4.1 <fft-grid-padding> . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.2.4.1.1 <x-fraction> . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.4.1.2 <y-fraction> . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.4.1.3 <z-fraction> . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.4.2 <use-intermediate-disk-storage> . . . . . . . . . . . . . . . . . 64
4.2.4.3 <vp-vs-ratio> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.2.4.4 <vp-vs-ratio-from-wells> . . . . . . . . . . . . . . . . . . . . . . 64
4.2.4.5 <maximum-relative-thickness-difference> . . . . . . . . . . . 64
4.2.4.6 <frequency-band> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.6.1 <low-cut> . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.6.2 <high-cut> . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.7 <energy-threshold> . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.8 <wavelet-tapering-length> . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.9 <minimum-relative-wavelet-amplitude> . . . . . . . . . . . . . . 65
4.2.4.10 <maximum-wavelet-shift> . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.11 <minimum-sampling-density> . . . . . . . . . . . . . . . . . . . . . 65
4.2.4.12 <minimum-horizontal-resolution> . . . . . . . . . . . . . . . . . 65
4.2.4.13 <white-noise-component> . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.14 <reflection-matrix> . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.15 <kriging-data-limit> . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.16 <debug-level> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.17 <smooth-kriged-parameters> . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.18 <rms-panel-mode> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.19 <guard-zone> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.20 <3d-wavelet-tuning-factor> . . . . . . . . . . . . . . . . . . . . . 66
4.2.4.21 <gradient-smoothing-range> . . . . . . . . . . . . . . . . . . . . . 67
4.2.4.22 <estimate-well-gradient-from-seismic> . . . . . . . . . . . . . 67
4.3 <survey> (necessary) . . . . . . . . . . . . . . . . . . . . . 67
4.3.1 <angular-correlation>. . . . . . . . . . . . . . . . . 67
4.3.2 <segy-start-time> . . . . . . . . . . . . . . . . . . 67
4.3.3 <angle-gather> (necessary). . . . . . . . . . . . . . . . 67
4.3.3.1 <offset-angle> (necessary) . . . . . . . . . . . . . . . . . . . . . . . 67
CRAVA User Manual version 1.2 9
4.3.3.2 <seismic-data> (necessary) . . . . . . . . . . . . . . . . . . . . . . . 67
4.3.3.2.1 <file-name> (necessary) . . . . . . . . . . . . . . . . . . . . 67
4.3.3.2.2 <start-time> . . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3 <segy-format> . . . . . . . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3.1 <standard-format> . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3.2 <location-x> . . . . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3.3 <location-y> . . . . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3.4 <location-il> . . . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3.5 <location-xl> . . . . . . . . . . . . . . . . . . . . 68
4.3.3.2.3.6 <bypass-coordinate-scaling> . . . . . . . . . . . 68
4.3.3.2.3.7 <location-scaling-coefficient> . . . . . . . . . 68
4.3.3.2.4 <type> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3 <wavelet> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.1 <file-name> . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.2 <ricker> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.3 <scale> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.4 <estimate-scale> . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.5 <local-wavelet> . . . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.5.1 <shift-file> . . . . . . . . . . . . . . . . . . . . . 69
4.3.3.3.5.2 <scale-file> . . . . . . . . . . . . . . . . . . . . . 70
4.3.3.3.5.3 <estimate-shift> . . . . . . . . . . . . . . . . . . 70
4.3.3.3.5.4 <estimate-scale> . . . . . . . . . . . . . . . . . . 70
4.3.3.4 <wavelet-3d> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3.3.4.1 <file-name> . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.3.3.4.2 <processing-factor-file-name> . . . . . . . . . . . . . 70
4.3.3.4.3 <propagation-factor-file-name> . . . . . . . . . . . . . 70
4.3.3.4.4 <stretch-factor> . . . . . . . . . . . . . . . . . . . . . . 71
4.3.3.4.5 <estimation-range-x-direction> . . . . . . . . . . . . . 71
4.3.3.4.6 <estimation-range-y-direction> . . . . . . . . . . . . . 71
4.3.3.5 <match-energies> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.3.6 <signal-to-noise-ratio> . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.3.7 <local-noise-scaled> . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.3.8 <estimate-local-noise> . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.4 <wavelet-estimation-interval> . . . . . . . . . . . . . 71
4.3.4.1 <top-surface-file> . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.4.2 <base-surface-file> . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.5 <time-gradient-settings> . . . . . . . . . . . . . . . 72
4.3.5.1 <distance> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
CRAVA User Manual version 1.2 10
4.3.5.2 <sigma> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4 <well-data> . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4.1 <log-names> . . . . . . . . . . . . . . . . . . . . . 72
4.4.1.1 <time> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4.1.2 <vp> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.4.1.3 <dt> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.1.4 <vs> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.1.5 <dts> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.1.6 <density> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.1.7 <facies> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.2 <well> . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.2.1 <file-name> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.4.2.2 <use-for-wavelet-estimation> . . . . . . . . . . . . . . . . . . . 73
4.4.2.3 <use-for-background-trend> . . . . . . . . . . . . . . . . . . . . . 73
4.4.2.4 <use-for-facies-probabilities> . . . . . . . . . . . . . . . . . . 73
4.4.2.5 <synthetic-vs-log> . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.2.6 <filter-elastic-logs> . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.2.7 <optimize-position> . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.2.7.1 <angle> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.2.7.2 <weight> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.3 <high-cut-seismic-resolution> . . . . . . . . . . . . . 74
4.4.4 <allowed-parameter-values> . . . . . . . . . . . . . . 74
4.4.4.1 <minimum-vp> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4.4.2 <maximum-vp> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.3 <minimum-vs> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.4 <maximum-vs> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.5 <minimum-density> . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.6 <maximum-density> . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.7 <minimum-variance-vp> . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.8 <maximum-variance-vp> . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.9 <minimum-variance-vs> . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.10 <maximum-variance-vs> . . . . . . . . . . . . . . . . . . . . . . . . 75
4.4.4.11 <minimum-variance-density> . . . . . . . . . . . . . . . . . . . . . 76
4.4.4.12 <maximum-variance-density> . . . . . . . . . . . . . . . . . . . . . 76
4.4.4.13 <minimum-vp-vs-ratio> . . . . . . . . . . . . . . . . . . . . . . . . 76
4.4.4.14 <maximum-vp-vs-ratio> . . . . . . . . . . . . . . . . . . . . . . . . 76
4.4.5 <maximum-deviation-angle> . . . . . . . . . . . . . . . 76
4.4.6 <maximum-rank-correlation> . . . . . . . . . . . . . . 76
CRAVA User Manual version 1.2 11
4.4.7 <maximum-merge-distance> . . . . . . . . . . . . . . . 76
4.4.8 <maximum-offset> . . . . . . . . . . . . . . . . . . . 76
4.4.9 <maximum-shift> . . . . . . . . . . . . . . . . . . . 76
4.4.10 <well-move-data-interval> . . . . . . . . . . . . . . . 77
4.4.10.1 <top-surface-file> . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.4.10.2 <base-surface-file> . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5 <prior-model> . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.1 <background>. . . . . . . . . . . . . . . . . . . . . 77
4.5.1.1 <ai-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.1.2 <si-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.1.3 <vp-vs-ratio-file> . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.1.4 <vp-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
4.5.1.5 <vs-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.6 <density-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.7 <vp-constant> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.8 <vs-constant> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.9 <density-constant> . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.10 <velocity-field> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.11 <lateral-correlation> . . . . . . . . . . . . . . . . . . . . . . . . 78
4.5.1.12 <high-cut-background-modelling> . . . . . . . . . . . . . . . . . 78
4.5.2 <earth-model> . . . . . . . . . . . . . . . . . . . . 79
4.5.2.1 <vp-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2.2 <vs-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2.3 <density-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2.4 <ai-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2.5 <si-file> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2.6 <vp-vs-ratio-file> . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.3 <local-wavelet> . . . . . . . . . . . . . . . . . . . 79
4.5.3.1 <lateral-correlation> . . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.4 <lateral-correlation>. . . . . . . . . . . . . . . . . 79
4.5.5 <temporal-correlation> . . . . . . . . . . . . . . . . 80
4.5.6 <parameter-correlation> . . . . . . . . . . . . . . . . 80
4.5.7 <correlation-direction> . . . . . . . . . . . . . . . . 80
4.5.8 <facies-probabilities> . . . . . . . . . . . . . . . . 80
4.5.8.1 <use-vs> . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.5.8.2 <use-prediction> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.5.8.3 <use-absolute-elastic-parameters> . . . . . . . . . . . . . . . . 80
4.5.8.4 <estimation-interval> . . . . . . . . . . . . . . . . . . . . . . . . 80
CRAVA User Manual version 1.2 12
4.5.8.4.1 <top-surface-file> . . . . . . . . . . . . . . . . . . . . . 81
4.5.8.4.2 <base-surface-file> . . . . . . . . . . . . . . . . . . . . 81
4.5.8.5 <prior-probabilities> . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5.8.5.1 <facies> . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5.8.5.1.1 <name> . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5.8.5.1.2 <probability> . . . . . . . . . . . . . . . . . . . . 81
4.5.8.5.1.3 <probability-cube> . . . . . . . . . . . . . . . . . 81
4.5.8.6 <uncertainty-level> . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.6 Variogram . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.1 <variogram-type> . . . . . . . . . . . . . . . . . . . 82
4.6.2 <angle> . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.3 <range> . . . . . . . . . . . . . . . . . . . . . . . 82
4.6.4 <subrange> . . . . . . . . . . . . . . . . . . . . . 82
4.6.5 <power> . . . . . . . . . . . . . . . . . . . . . . . 82
A Sample model le. . . . . . . . . . . . . . . . . . . . . . . . . 83
B Test suite overview . . . . . . . . . . . . . . . . . . . . . . . . 86
C Release notes . . . . . . . . . . . . . . . . . . . . . . . . . . 88
C.1 Changes from v1.1 to v1.2 . . . . . . . . . . . . . . . . . . . 88
C.2 Changes from v1.0 to v1.1 . . . . . . . . . . . . . . . . . . . 88
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
CRAVA User Manual version 1.2 13
1 Model
1.1 Introduction
Seismic inversion has traditionally been treated as a deterministic problem. However, there are
several uncertain aspects: There is noise in the seismic amplitude data, and the frequency resolu-
tion is limited. This means that neither high nor low frequencies can be resolved from the seismic
amplitude data alone. Using a geostatistical approach to the problem of seismic inversion, the
uncertainty may be treated in a consistent and robust way.
The CRAVA program uses the Bayesian linearised AVO inversion method of Buland et al. (2003)
to take the uncertainty in seismic amplitude data into account. Since we only use amplitude data,
we will from here on use the term seismic data for these. The seismic data are described using
multi-normal distributions, and modelled as the seismic response of the earth model plus an
error term. The earth model and error term are modelled as multi-normal distributions in which
spatial coupling is imposed by correlation functions. Using a Bayesian setting, prior models for
the earth and error terms are set up based on prior knowledge obtained from well logs, and the
process of seismic inversion is reduced to that of nding a posterior distribution for the earth
given the seismic data. The linearised relationship between the model parameters and the AVO
data, makes it possible to obtain the posterior distribution analytically.
The posterior distribution for earth model parameters V
p
(pressure-wave velocity), V
s
(shear-
wave velocity), and (density), gives a laterally consistent seismic inversion. The lateral corre-
lation can follow the stratigraphy of the inversion interval by following the top and/or base of
the inversion volume, but can also be specied independently using a correlation surface. As a
consequence of the spatial coupling, the solution in each location depends on the solutions in
all other locations. From the posterior distribution the best estimate of the model parameters
and a corresponding uncertainty can be extracted. Moreover, since the distribution is Gaussian,
kriging can be used to match the well data, and the posterior covariance can be computed. This
spreads full frequency information in an area around the wells. Full frequency realizations can be
generated by sampling from the posterior distribution. A set of such realizations represents the
uncertainty of the inversion.
1.2 AVO
Amplitude versus offset (AVO) inversion can be used to extract information about the elastic sub-
surface parameters from the angle dependency in the reectivity, see e.g., Buland et al. (1996);
Hampson and Russell (1990); Lrtzer and Berkhout (1993); Pan et al. (1994). In practice, and espe-
cially for 3-D surveys, linearised AVO inversion is attractive since it can be performed with use
of moderate computer resources. Prior to a linearised AVO inversion, the seismic data must be
processed to remove nonlinear relations between the model parameters and the seismic response.
Important steps in the processing are the removal of the move-out, multiples, and the effects of
geometrical spreading and absorption. The seismic data should be pre-stack migrated, such that
dip related effects are removed. After pre-stack migration, it is reasonable to assume that each sin-
gle bin-gather can be regarded as the response of a local 1-Dearth model. The benets of pre-stack
migration before AVO analysis are discussed in Brown (1992); Buland and Landr (2001); Mosher
et al. (1996). It is further assumed that wave mode conversions, interbed multiples and anisotropy
CRAVA User Manual version 1.2 14
effects can be neglected after processing. Ideally, the pre-stack gathers are also transformed from
offsets to reection angles. Offset gathers are often close to angle gathers, so we can use these, but
if the inversion area is thick, this will give more noise in the CRAVA model.
1.3 Seismic model
The seismic response of an isotropic, elastic mediumis completely described by the three material
parameters {V
p
(x, t), V
s
(x, t), (x, t)}, where the vector x gives the lateral position (x,y), and t is
the vertical seismic travel time.
The weak contrast approximation by Aki and Richards (1980), relates the seismic reection coef-
cients c(x, t, ) to the elastic medium, and is a linearization of the Zoeppritz equations. A con-
tinuous version of this approximation is given by Stolt and Weglein (1985):
c(x, t, ) = a
Vp
()

t
lnV
p
(x, t)
+ a
Vs
(x, t, )

t
lnV
s
(x, t)
+ a

(x, t, )

t
ln(x, t),
(1.1)
where is the PP reection angle, and
a
Vp
() =
1
2
_
1 + tan
2

_
,
a
Vs
(x, t, ) = 4
V
2
s
(x, t)
V
2
p
(x, t)
sin
2
, (1.2)
a

(x, t, ) =
1
2
_
1 4
V
2
s
(x, t)
V
2
p
(x, t)
sin
2

_
for PP reections, and
a
Vp
() = 0
a
Vs
(x, t, ) = 2
sin
cos
_
V
2
s
(x, t)
V
2
p
(x, t)
sin
2

V
s
(x, t)
V
p
(x, t)
cos cos
_
, (1.3)
a

(x, t, ) =
sin
cos
_

1
2
+
V
2
s
(x, t)
V
2
p
(x, t)
sin
2
+
V
s
(x, t)
V
p
(x, t)
cos cos
_
,
for PS reections. Here, is the PS reection angle, given by sin = (V
s
/V
p
) sin. These equations
are linearised by replacing the ratio V
s
(x, t)/V
p
(x, t) with a constant value

V
p
/

V
s
when computing
the coefcients.
The seismic data are represented by the convolutional model
d
obs
(x, t, ) =
_
w(, ) c(x, t , ) d + e(x, t, ), (1.4)
where w is the wavelet, and e is an angle and location dependent error term. The integral is the
synthetic seismic. The wavelet can be angle dependent, and can vary laterally according to scale
and shift maps. The wavelet is assumed to be stationary within a limited target window.
The signal-to-noise ratio is dened as the ratio of the energy of the data to the energy of the noise
as given in Equation 1.4, that is,
S/N =
_
w c
2
+e
2
_
/e
2
, (1.5)
where the operator * denotes the convolution. Since the error is independent of the synthetic
seismic, the energy form the synthetic seismic and the noise can simply be added. Note that
there also exists another denition of the S/N ratio where the noise energy is not included in
the enumerator.
CRAVA User Manual version 1.2 15
1.3.1 Convolution with 3D wavelet
The seismic data can also be represented as a convolution in three dimensions
d
obs
(x, t, ) =
_
w(x, , ) c(x , t , ) dd + e(x, t, ), (1.6)
where w now denotes a 3D wavelet. The 3D wavelet is derived from the point-spread function
(PSF) which acts as a lter on the reectivity cube to mimic the imaging process of pre-stack depth
migration. For a thorough description on howthe PSF is calculated with ray-tracing by identifying
the illumination vectors, see Lecomte (2008). For a more physical interpretation of the PSF, the
parametrisation in the wavenumber (Fourier) domain is done in terms of spherical coordinates.
The relationship between these coordinates and the spatial frequencies k = (k
x
, k
y
, k
z
) for a point
P is given by
k
x
= r cos() sin(), k
y
= r sin() sin(), k
z
= r cos(), (1.7)
where r is the radial distance fromorigo to P, is the azimuth angle between the line fromorigo to
P projected into the (k
x
, k
y
)-plane and the k
x
-axis and is the dip angle dened as the inclination
angle between the line from origo to P and the upward pointing k
z
-axis. varies between 0

and
360

while varies between 0

and 90

implying that only upwards scattering reections are

considered, and not turning waves.
Dening the temporal frequency = V
0
r(2 cos )
1
where V
0
is the average velocity for the region
of interest and cos represents a stretch factor due to reection angle, the model for the point-
spread function, f, is given by

f(r, , ; ) =
1
(, ; )
2
(, , ; ) w
0
(; ) , (1.8)
where the tilde denotes Fourier transform, w
0
(; ) is the 1D pulse and functions
1
(, ; ) and

2
(, , ; ) = exp(||

H(, ; )) are frequency independent and frequency dependent pro-

cessing factors respectively.
Since the point-spread function is dened in the depth domain, and the 3D wavelet in expression
(1.6) is in time domain, a relation between time and depth is needed. In the region of interest, a
target centre is dened at depth Z
0
. A reference time surface T
0
(x, y) corresponds to this depth.
For a given time the corresponding depth is . Let T

(x, y) be the time to a point, and Z

(x, y)
the depth to the same point, the time-depth relation is:
T

(x, y) = T
0
(x, y) +
2
V
0
(Z

(x, y) Z
0
) (1.9)
1.4 Statistical model
The elastic parameters V
p
(x, t), V
s
(x, t), and (x, t) are assumed to be log-normal random elds.
This means that the distribution m(x, t) = [lnV
p
(x, t), lnV
s
(x, t), ln(x, t)]
T
is multi-normal or
multi-Gaussian, that is,
m(x, t) N
_

m
(x, t),
m
(x
1
, t
1
; x
2
, t
2
)
_
, (1.10)
where
m
(x, t) are the expectations of m(x, t) and
m
(x
1
, t
1
; x
2
, t
2
) gives the covariance struc-
ture. We assume that the covariance function is stationary and homogeneous (i.e., translationally
invariant), and can be factorised as

m
(x
1
, t
1
; x
2
, t
2
) =
0,m

m
()
m
(), (1.11)
where
m
() and
m
() are correlation functions depending on the lateral and temporal distances
= x
2
x
1
and = |t
2
t
1
|, respectively, and
0,m
is a 3 3 matrix of the variances and
CRAVA User Manual version 1.2 16
covariances of lnV
p
, lnV
s
and ln. Any covariance structure giving a positive denite
m
may
be used.
If we let m and d
obs
be discrete representations of m(x, t) and d
obs
(x, t, ) in a time interval,
Equation (1.4) may be written in matrix notation as
d
obs
= WADm+e (1.12)
= Gm+e (1.13)
where Wis the matrix representation of the wavelets, A is a matrix encompassing discrete rep-
resentations of the coefcients a
V p
, a
Vs
, and a

, D is a differential matrix and G = WAD. The

error matrix e is a time discretization of the error vector e(x, t) = [e(x, t,
1
), . . . , e(x, t,
n

)]
T
and
is assumed to be zero-mean coloured Gaussian noise, that is,
e(x, t) N
n

_
0,
e
(x
1
, t
1
; x
2
, t
2
)
_
. (1.14)
The covariance of the error vector is

e
(x
1
, t
1
; x
2
, t
2
) =
0,e

e
()
e
(), (1.15)
where
0,e
is an n

covariance matrix containing the noise variances for the different reec-
tion angles on the diagonal, and the covariances between the angles off the diagonal. Furthermore,

e
() and
e
() are lateral and temporal correlation functions, similar to those given for m(x, t) in
Equation (1.11).
Since the relationship between the reection coefcients and the elastic parameters given in Equa-
tion (1.1) is linear, and the elastic parameters are assumed Gaussian distributed, the reection
coefcients become Gaussian. Moreover, since the convolution is a linear operation and we have
assumed a Gaussian error model, the seismic data given in Equation (1.4) are also Gaussian dis-
tributed.
For the time-discretized seismic data d
obs
, this gives us the multi-normal distribution
d
obs
N
n
d
_

d
,
d
_
, (1.16)
where

d
= G
m
, (1.17)

d
= G
m
G
T
+
e
. (1.18)
where all vectors and matrices are time-discretized.
This means that the simultaneous distribution for mand d
obs
is Gaussian, and that the distribu-
tion for m given d
obs
can be obtained analytically using standard theory for Gaussian distribu-
tions:

m|d
obs
=
m
+
m
G
T

1
d
(d
obs

d
) (1.19)

m|d
obs
=
m

m
G
T

1
d
G
m
, (1.20)
where
d
is the expected observation, that is, the seismic response of
m
, and
d,m
is the covari-
ance matrix between logarithmic parameters and observations. See Buland and Omre (2003) for a
detailed description on how to compute these.
The computations given in Equations (1.19) and (1.20) involves the inverse of
d
. Given an inver-
sion volume with n cells, this matrix has n
2

n
2
elements, and for any reasonably sized volumes,
CRAVA User Manual version 1.2 17
inverting this matrix is forbiddingly time consuming. However, the covariance function for a
homogeneously correlated spatial variable is diagonalised by a 3D Fourier transform (Christakos
(1992)), and in this domain the inversion problemcan be solved independently for each frequency
component. This reduces the complexity of the computations dramatically, and the calculation
time becomes O(nlog n). This is illustrated in Figure 1.1. Details can be found in Buland et al.
(2003)
Figure 1.1. The problem is transformed to the Fourier domain, solved in this domain, and back-transformed
to time domain. This reduces the problem from a O(n
2.x
) to a O(nlog n) process.
This seems to require that d is stationary, which would imply that the wavelet must be the same
everywhere. However, we can divide out the wavelet from Equation 1.12 to obtain
d

= ADm+e

, (1.21)
where d

is the data divided by the wavelet, and e

is the error divided by the wavelet. The details

of how we do this division is given in Section 2.6.1. Note that we now assume that the noise after
division, e

is stationary. Since we assume that a seismic response only depends on the reections
in that trace, this division can be done trace by trace. This assumption relies on a rather smooth
seismic response, so the lateral variations in the wavelet should be smooth. We have chosen to
restrict the local wavelet changes to only allow local temporal shift and amplitude scaling.
We can also work around the stationary noise assumption, and allow e

to vary laterally. This is

done by utilising the nature of a Bayesian solution, which always is a tradeoff between the prior
and the posterior with noise free data. By nding the posterior for a minimal noise, we can then
interpolate between this solution and the prior to nd an appropriate solution for the local noise
level. When doing this, we ignore the lateral correlation, so we require constant noise level in each
trace, since the conditional correlations inside a trace are much stronger than between traces. See
Section 2.6.2 for details.
1.4.1 Facies probabilities
To calculate facies probabilities from inverted elastic parameters we must rst establish a link
between the elastic parameters from inversion and facies. There are several ways to do this, and
we have listed the four most realistic in Table 1.1.
First, we may establish the link using a rock physics models. This way we avoid any alignment
problems, but we get no frequency control and get to use no correlation information from the
CRAVA User Manual version 1.2 18
Table 1.1. Different methods for establishing a relation between elastic parameters and facies.
Approach Frequency Inversion Alignment Predictions
control correlation
Rock physics No No Yes Optimistic
Low-pass ltered elastic well logs Yes No Yes Optimistic
Inversion Yes Yes No Pessimistic
Parameter ltered elastic well logs Yes Yes Yes Realistic
Figure 1.2. Acoustic impedance residuals calculated from blocked raw logs plotted against corresponding
V
p
/V
s
residuals (left), the same plot, but with logs high-cut frequency ltered to 40Hz (middle), and the same
logs but ltered using frequency and correlation information from inversion (right).
inversion. A prediction obtained this way tends to be too optimistic, but if there are no well logs
available, this is our only option.
Second, we may lter the elastic well logs using a low-pass lter, and use these ltered logs
and facies logs to make a density estimation of p(
m|d
obs
|f), where f is the facies. This gives us
frequency control and proper alignment, but again there are no correlation information from the
inversion included, and the predictions become to optimistic. The frequency ltering of well logs
is illustrated in Figure 1.2.
Third, we may set up the probability density for an elastic responds given the facies directly from
the inverted elastic parameters. This way we get both frequency control and correlation informa-
tion included, but we get no alignment information, and the predictions become to pessimistic.
Finally, we may establish the density using elastic parameters from well logs, but lter these
using frequency and correlation information obtained from the inversion. Since we only use well
logs to establish the density estimates, we get no alignment problems, and in total, this gives
us realistic facies predictions. How the inversion-based ltering differs from the pure frequency-
based ltering mentioned above is illustrated in Figure 1.2.
The link between facies and elastic parameters that we eventually are interested in is p(f
i
|
m|d,i
),
where f
i
is the facies at location i and
m|d,i
is the inversion result at the same location. To estab-
lish this link, we must rst establish a link between the well logs mand the expectation given the
seismic data
m|d
obs
. We get this by combining Equation 1.13 and Equation 1.19:

m|d
obs
=
m
+
m
G
T

1
d
(d
obs

d
)
=
m
+
m
G
T

1
d
(Gm+e G
m
)
=
m
+F(m
m
) +e

, (1.22)
CRAVA User Manual version 1.2 19
Figure 1.3. To the left, the posterior sand probability calculated without undened facies. In the middle, the
sand probability when undened facies is introduced. To the right, the posterior probability for undened
facies.
where
F =
m
G
T

1
d
G (1.23)
e

N(0,
e
) (1.24)

e
=
m
G
T

1
d

e

1
d
G
m
. (1.25)
The operator F is used to lter the well logs and obtain an estimate of the expected inversion
values m

. For each facies, we then do a density estimation of p(

m|d
obs
|f) for each possible
facies value f using m

and facies logs. The density estimation is done using a kernel smoothing
approach, and by using the distribution of e

as our kernel, we get an unbiased estimate of this

distribution.
Finally, we nd the facies probability:
p(f = j|
m|d
obs
) =
p(
m|d
obs
|f = j)p(f = j)

i
p(
m|d
obs
|f = i)p(f = i)
, (1.26)
where p(f = i) is the prior probability of facies i. This probability is then computed for each facies
and each cell in the grid, with
m|d
obs
given by the inversion results.
Far away from wells, the estimates will not be reliable, and we introduce an undened facies to
show such areas. Denoting the likelihood of this undened facies p(u), the facies probabilities are
now calculated as
p(f = j|
m|d
obs
) =
p(
m|d
obs
|f = j)p(f = j)

i
p(
m|d
obs
|f = i)p(f = i) + p(u)
, (1.27)
where p(u) is uniform over the area, and low compared to the likelihood for facies when we
are close to data. In Figure 1.3 the effect of the undened facies in a reservoir consisting of sand
and shale is illustrated. The three gures show cross plots of well observations of against V
p
combined with a probability map shown as a Colo coded map.
In the left gure, we showthe probability of sand before the undened facies has been introduced.
Note howcombinations of and V
p
which are far away fromany well observations, as for instance
(V
p
, ) = (2000m/s, 2.25g/cm
3
), may still lead to a facies prediction of sand equal to one. This is
not realistic.
In the middle gure, the undened facies has been introduced, and whenever we get far away
from combinations of and V
p
for which we have no well observations, the probability for sand
now decreases to zero. This does not mean that there is no chance of nding sand at the cur-
rent spot, only that we have no data support to tell us what facies we might nd. Similarly, the
probability of nding shale will also be zero.
CRAVA User Manual version 1.2 20
In the right gure, we showthe probability of the undened facies. This is zero around well obser-
vations, and gradually increases to one as the distance to observations increase. The probability
of shale may be extracted from these gures since p(sand) + p(shale) + p(undened) = 1.
CRAVA User Manual version 1.2 21
2 Implementation
Whereas the general model was explained in Chapter 1 we explain a bit more of the actual imple-
mentation details here.
The estimation routines implemented in CRAVA are based on straightforward and commonly
used techniques. This gives fast and robust estimation, although we may run into problems if the
number of data points is too small, or the data quality is too low. The quality of an estimation
result is never better than the quality of the data it is based on.
2.1 Estimating optimal well location
The positioning uncertainty between well data and seismic data is often signicant. To overcome
this, the well may be moved to the location with maximum correlation between the seismic data
and the reection coefcients calculated from the well data. The relation between the seismic
data and reection coefcients is linear; so linear covariance is a good measure. The optimal well
location is found by searching for the location with highest covariance in a lateral neighbourhood
around the original well location, where the well is allowed to be shifted vertically in each target
position. The moving of wells is triggered by a command in the model le, and it is done prior to
the estimation of wavelets, noise, correlations and background model.
2.2 Estimating the prior model
The prior model for the Bayesian inversion is dened in equation (1.10), and consists of the ex-
pectations of the elastic parameters V
p
, V
s
, and collected in the vector
m
, and their spatial
correlation structure collected in the covariance matrix
m
. These expectations and covariances
must be given prior values before the inversion.
2.2.1 Background model
The expectation
m
is usually referred to as the background model. As the seismic data do not
contain information about low frequencies, a background model is built to set the appropriate
levels for the elastic parameters in the inversion volume. To identify this level, we can plot the
frequency content of the seismic traces in the available wells, and identify lowest frequency for
which seismic data contains enough energy to carry information.
In Figure 2.1, we have plotted the frequency content in the seismic data in two different wells. The
green curve gives the frequency content in the near stack and the blue curve gives the frequency
content in the far stack. These plots show that the seismic data contain little energy below 56Hz,
and the purpose of the background model is to ll this void.
The estimation of the background model is made in two steps. First, we estimate a depth trend for
the entire volume, and then we interpolate well logs into this volume using kriging. The estima-
tion will by default contain information up to 6Hz, but this high-cut limit can be adjusted using
the <high-cut-background-modelling> keyword.
When identifying the depth trend, it is important that the wells are appropriately aligned. The
alignment is dened by the time interval surfaces specied as input, or alternatively, the correla-
tion direction surface. It is important that the alignment reects the correlation structure (deposi-
CRAVA User Manual version 1.2 22
Figure 2.1. The frequency content of the seismic traces in two different wells. The frequency content of the
near and far stacks are shown as green and blue curves respectively.
Figure 2.2. Well logs aligned according to true time scale (left) and according to stratigraphic depth (right).
tion/compaction), and if the time surfaces are either eroding or on-lapped, one should consider
specifying the correlation direction separately using the <correlation-direction> keyword.
In Figure 2.2, we show two well logs aligned according to deposition and according to the true
time scale. Evidently, an incorrect trend will be identied if the true vertical depth is used. The
size of the error will depend on the stratigraphy.
Assuming properly aligned wells, the trend extraction starts by calculating an average log value
for each layer. This average is calculated for the V
p
, V
s
, and well logs and is based on all avail-
able wells. The estimation uses a piecewise linear regression, rather than the more straightforward
arithmetic mean or moving average, as these measures are sensitive to the amount of data avail-
able. The piecewise regression has the additional advantage that it can give trend estimates also
outside the interval for which we have data available.
For the linear regression we require a minimum of 10 data points behind each estimate. In addi-
tion, we require that the minimum number of data points must also be at least 5*N
wells
. This way
we ensure that data points from different time samples are always included. Alternatively, the
regression would reduce to an arithmetic mean whenever there are 10 or more wells available. If
we enter a region with no data points available at all, the minimum requirements are doubled.
CRAVA User Manual version 1.2 23
Figure 2.3. Well log values plotted against grid layer number for V
p
(left) and (right). The blue circles show
log values, the green curve is a piecewise linear regression of the these values, and the red curve is the
regression values ltered to 6Hz.
Figure 2.4. V
p
depth trend (left) and nal background model (right). Well logs of V
p
, high-cut ltered to 6Hz,
are shown for comparison.
To get the right frequency content in the depth trends, the regression values are eventually fre-
quency ltered to 6Hz.
The trend extraction process is illustrated in Figure 2.3 for the V
p
and logs of a eld with six
wells. Note that the plots are oriented with layers as abscissa and log values as ordinate. The blue
circles represent log values from any wells, the green curve is the piecewise linear regression of
these values, and the red curve is the frequency ltered log that will be used as a depth trend.
Note that the green curve is slightly erratic, especially, as we enter the region (below reservoir)
where there are no data points available. This shift, which is clearly observed for the density,
arises as we stabilise the estimate by requiring twice as many data points behind each estimate.
When the inversion volume has been lled with the depth trend, we interpolate it with 6Hz
ltered well logs, to ensure that the background model will match in wells. A cross section of the
resulting background model for V
p
is illustrated in the right part of Figure 2.4. To the left is the
corresponding depth trend. For comparison the well logs of V
p
has plotted in both illustrations.
Note how the wells inuence the volume in a region around the well.
Ideally, the background model should be as smooth as possible, and a Gaussian variogram model
with relatively long ranges may seem an obvious choice. This model is too smooth, however, and
should be omitted as it often give parameter over- and undershooting away from wells.
CRAVA User Manual version 1.2 24
2.2.2 Covariance
Since we model the covariance structure as separable, we have collapsed the full time dependent
covariances between parameters into one parameter covariance matrix
0,m
, a lateral correlation
vector
m
(), and a temporal correlation vector
m
().
We estimate the correlations by rst blocking the wells into the grid, and then do standard corre-
lation estimation using
Cov(X, Y ) =

(x
i
x)(y
j
y)

n 1
, with X, Y {lnV
p
, lnV
s
, ln} (2.1)
The parameter covariance matrix is simply estimated by using the covariances at time lag 0. In
Figure 2.5, we showcross plots of the parameter residuals (x
i
x) for a sample eld. The depicted
distributions look similar to bivariate normal distributions, which supports the normal distribu-
tion assumptions made in Equation 1.10. If there are no V
s
logs available, the prior V
s
variance
will be set equal to twice the V
p
variance, and their covariance will be set equal to zero.
The temporal correlation is estimated from the remaining lags in the well logs as depicted in
Figure 2.6. The temporal correlation will be a weighted average of the estimates made for all
three elastic parameters.
While the covariance matrix and the temporal correlation can be readily estimated fromwell data,
this is not the case for the lateral correlation, unless there are a large number of wells available.
The lateral correlation is therefore normally chosen parametric. There is an option in CRAVA to
estimate the lateral correlation from seismic data, but these estimates are not made relative to
stratigraphy and tend to grossly underestimate the correlation. Using a parametric correlation
function is therefore encouraged. In Figure 2.7, we have depicted an exponential correlation func-
tion and the lateral correlation structure this kind of function gives rise to.
2.2.3 Likelihood model
As with the prior model for the elastic parameters, we have also collapsed the full time dependent
error covariance matrix into a noise covariance matrix
0,e
, a lateral correlation vector
e
(), and
a temporal correlation vector
e
().
The lateral correlation is difcult to estimate and is chosen equal to that for the elastic param-
eters, that is, we use
e
() =
m
(). The temporal correlation is partly estimated from wavelet
derivatives and partly white noise. By default, a 10% white noise fraction is assumed.
For the noise covariance matrix, the noise for a single angle gather can be either specied in the
model le using the <signal-to-noise> keyword or it can be estimated. Anoise estimate is found
by generating synthetic seismic data (see next section) using the wavelet optimally shifted in each
well, and subtracting this from the seismic data. The remaining part is assumed to be noise, and
we measure the noise energy from this.
The correlation between the noise in different angle stacks is hard to estimate and is therefore
chosen parametric. Typically, an exponential correlation functions with a range of 10

is used. In
most cases this implies that the noise in the angle stacks are treated as independent of each other.
2.3 Estimating wavelets
The implemented wavelet estimation uses the approach of spectral division, see White (1984). In
this approach an estimate of the cross-correlation between data and reection coefcients, and
an estimate of the auto-correlation of reection coefcients are used to estimate the wavelet. The
methodology requires that the reection coefcients are known, thus wavelets are estimated at
well locations. The cross-correlation between data and reection coefcients is found by convolv-
CRAVA User Manual version 1.2 25
Figure 2.5. Cross plots of logarithmic parameter residuals. From such plots the parameter correlations may
be estimated.
Figure 2.6. We obtain the temporal covariance by measuring the covariance between all pairs of points in
the well log (left). The resulting temporal correlation function (right).
Figure 2.7. Parametric lateral correlations. A two-dimensional exponential correlation function (left). The
lateral correlation structure resulting from an anisotropic exponential correlation function having an azimuth
of 45

degrees (right).
CRAVA User Manual version 1.2 26
Figure 2.8. The local scale and shift maps involved when using local wavelets
ing the data with the reection coefcients, and tapering the result. The auto-correlation of the re-
ection coefcients are found similarly by convolving the reection coefcients with themselves
and then applying a taper to the result. The tapering is performed in order to avoid spurious
correlations at large lags.
Using the standard convolutional relation for seismic data,
d = w c +e, (2.2)
where d is the seismic amplitude data, w is the wavelet, c the reection coefcients, and e is the
noise. We see that convolving the data with reection-coefcients, transforming to the Fourier
domain, and take the expectation we get
d() c() = w()|c()|
2
(2.3)
Note that the convolution has disappeared, and the equation can be solved for each frequency .
We recognise the left hand side as the spectre of the cross-correlation between data and reection
coefcients. And the left hand side as the wavelet multiplied with spectre of the auto-correlation
of the reection coefcients. This can be obtained by dividing the spectre of the cross-correlation
with the spectre of the auto-correlation.
Tapering of the estimated cross-correlation and auto-correlation is required in order to stabilise
the estimate. In Crava a Papoulis taper is used. Tapering is equivalent to a local smoothing in the
frequency domain, thus the resulting wavelet estimate will behave smoothly in Fourier domain.
We nd the optimal vertical shift for each well. The global wavelet is then found by taking the
arithmetic average of the zero-phase wavelets, weighted by the number of samples used from
each well.
When using local wavelets, we nd the optimal shift and/or scale of the global wavelet at each
well location. Optimal here means minimising the noise energy. We then use kriging to interpolate
this between wells, with a shift of 0 and a scale of 1 as the mean level outside the well control area.
This is illustrated in Figure 2.8.
Local noise is estimated using the local noise energies fromabove. We always use local shift when
estimating the noise, but only use local scale if it is used in the inversion. If local scale is used, the
noise is divided by this. Anoise scaling factor is then computed in each well, and kriged as above.
CRAVA User Manual version 1.2 27
2.4 Estimating 3D wavelet
The expression for the wavenumber representation of the point-spread function given in Equa-
tion 1.8 has only one unknown element, namely the 1D pulse w
0
(). The functions
1
and

H are
given as input, together with the average velocity V
0
. The elements needed for the conversion
from depth to time are also input to CRAVA. That is the reference depth Z
0
, and reference time
surface T
0
. The theory for the estimation of the 1D pulse is given in Georgsen et al. (2010a), and
with more details in Georgsen et al. (2010b).
As for the 1D wavelet, the pulse is estimated from wells. Using reection coefcients from well
logs, time gradients estimated from seismic data around wells and depth gradients computed
from time gradients by using the reference time surface and average velocity given, a matrix K
can be constructed forming a linear regression model for the seismic data as
d = Kw
0
+e. (2.4)
The least squares estimate for w
0
is
w
0
= (K

K)
1
G

d. (2.5)
2.5 Using FFT for inversion
As previously stated, Equation 1.12 separates when transformed into the Fourier domain. After
this transformation, the equation becomes

d(, k) = G() m(, k) +e(, k)) (2.6)

The tilde denotes the 3D Fourier transform, with temporal frequency , and lateral frequency
vector k = (k
x
, k
y
). Due to the separation, we now have a set of n small equations, where n
is the number of grid cells in the inversion volume. Everything is still normally distributed, so
the solution to this equation follows the pattern from Equation 1.19 and Equation 1.20. We still
must invert a data covariance matrix, but whereas this matrix had dimension (n n

)
2
before the
Fourier-transform, the matrix we must invert here is reduced to dimension n
2

, where n

is the
number of angle stacks. Since the time for a matrix inversion is almost cubic in size, it is much
faster to invert n of these small matrixes than the one large. After solving for m(, k), we do the
inverse transform of this to obtain the distribution for m. The same does of course hold when
we are using local wavelets that are divided out in advance, Equation 1.21. For full details, see
Buland et al. (2003).
2.6 A note one local wavelet and noise
As shown, even though the use of FFT-transformrequires stationarity, we are able to work around
this. Wavelets can be made local since these can be divided out before solving the equations, and
locally higher noise levels can be approximated by interpolating the low-noise solution and the
prior distribution.
2.6.1 Local wavelet - dividing out the wavelet
A simple division of data by wavelet can easily be done in the Fourier domain, where the convo-
lution reduces to a multiplication, and the division can be done one frequency at a time. However,
this is very unstable for frequencies where the wavelet is very weak or not present, and some sort
of stabilisation is needed.
In CRAVA this is done in two ways. First, we set an upper and lower cutoff frequency for the
wavelet, default set to 5 and 55 Hz. Furthermore, for frequencies that fall below10%of the average
amplitude, we set the amplitude to 10% of average before doing the division.
CRAVA User Manual version 1.2 28
2.6.2 Local noise
Local noise is implemented by rst nding the solution using the minimum noise level, to full
the stationarity requirements of the FFT algorithm. We then interpolate the values for each lo-
cations between the prior and this minimum noise posterior. When doing this interpolation, we
ignore correlation between locations. This is not a problem as long as the noise varies slowly and
smoothly.
For each location x the adjusted estimate
m|d
obs
(x), is found from the inversion result
m|d
obs
(x)
by a linear relation,

m|d
obs
(x) =
m
(x) +H
x
_

m|d
obs
(x)
m
(x)
_
, (2.7)
The matrix H
x
is a shrinkage matrix, i.e. the adjusted estimate is always closer to the prior mean
than the inversion result. The matrix H
x
depends on the local error variance
x
e
and error vari-
ance used in the inversion
0
e
.
To nd the shrinkage matrix we rst identify a matrix G
0
which maps the local prior distribution
to the local posterior distribution when it is observed with the noise
0
e
, that is,
d(x) = G
0
m(x) +e
0
,
where e
0
N
_
0,
0
e
_
. The inversion of this expression is a linear relation

m|d
obs
=
m
+P(
0
e
) (d
obs

d
) (2.8)
where P(
0
e
) =
m
G
T
0
_
G
0

m
G
0
T
+
0
e
_
1
.
We then dene the shrinkage matrix to be:
H
x
(
x
e
,
0
e
) = P(
x
e
)P(
0
e
)
1
. (2.9)
This removes the effect of the standard inversion and add the effect of the locally adapted inver-
sion. The matrix P(
0
e
) is not invertible, but since the local noise always is larger than the noise
in the inversion the product in expression 2.9 is always well dened.
2.7 Memory handling
Since the grids needed by CRAVA can become very large, we try to keep the number of grids
kept simultaneously in memory as small as possible. This implies that some allocated grids will
be used for more than one purpose. Both padded and unpadded grids are used in CRAVA, The
amount of memory needed by padded and unpadded grids are denoted s
p
and s
u
respectively.
CRAVA also has an option to use disk space for intermediate storage of grids. This will reduce
the memory consumption with a factor of at least 2 in realistic cases, but will also increase the
computation time by a factor of almost 3.
2.7.1 Grid allocation with all grids in memory
If intermediate disk storage is not used, the grid memory allocation will go as follows:
1. Background grids for Vp, Vs and density, 3 grids.
2. If the background model is to be estimated, another 3 grids are allocated for estimation, but
destroyed before any other allocations.
3. Seismic grids, n

. If well optimisation is used, this will come before background grids.

4. Possibly prior facies probability grids, indicated by I
p
, n
f
unpadded grids.
CRAVA User Manual version 1.2 29
5. Prior covariance, 6 grids.
6. If relative facies probabilities or local noise, a copy of background indicated by I
b
, 3 grids.
7. Peak: At this stage, CRAVA reaches its rst memory peak. Minimum memory in use is 10
grids, typical situation with three seismic grids and facies modelling requires 15 grids.
Memory usage: P
1
= (9 + n

+ I
b
3) s
p
+ I
p
n
f
s
u
.
8. The posterior distribution is computed into the background and prior covariance grids, and
seismic residuals are computed into the seismic grids. Thus, the inversion requires no extra
grids. (But we needed a copy of the background for local noise or facies.)
9. After the inversion, the seismic grids are released, taking us off peak down to a base level:
Memory usage: P
base
= (9 + I
b
3) s
p
+ I
p
n
f
s
u
.
10. If simulation is used:
a. Simulated grids are allocated, 3 grids.
b. If secondary elastic parameters are requested as output (AI, , etc.), indicated by I
s
, a
computation grid is allocated, 1 grid.
c. If kriging is used, indicated by I
k
, 1 unpadded grid, not concurrent with computation
grid.
11. Peak: New possible peak, since the number of grids now allocated may be larger than the
released seismic grids.
Memory usage: P
2
= (12 + I
b
3 + I
s
) s
p
+ (I
p
n
f
+ max(0, I
k
I
s
)) s
u
.
12. New release of grids, back to P
base
.
13. If facies probabilities:
a. 3D histograms of elastic parameters per facies are created, each of size 2MB, n
f
special
grids.
b. Facies probability grids are created, including for undened, n
f
+ 1 unpadded grids.
14. Peak: New possible peak, since the new memory allocated may be larger than the released
seismic grids and/or the simulation+computation/kriging grids.
Memory usage: P
3
= (9 + I
b
3) s
p
+ (I
p
n
f
+ n
f
+ 1) s
u
+ 2 n
f
.
15. Can now release all grids related to facies probability, memory down to 9 s
p
.
16. Eventual kriging of prediction allocates 1 unpadded grid.
17. Everything released.
The maximum memory usage is thus the largest of the actual peaks. The maximum number of
allocated padded grids will occur at either P
1
or P
2
, whereas the largest number of other grids
are allocated at P
3
.
CRAVA User Manual version 1.2 30
3 User guide
In this chapter, we describe how to build a CRAVA model le. The model le mainly follows the
XML format, but we also use the character # for commenting, meaning that the rest of the line
after such a character is read as comment. XML les are built with start and end tags, encapsulat-
ing either tags or values. All model les start with <crava>, and end with </crava>. An example
of a model le is given in Appendix A.
3.1 Basic inversion
Aprimary ability for CRAVAis to run simple rst-pass inversions. In this section, we describe how
to build a model le for a simple inversion. We focus on how to get the key information into the
program, whereas more detailed controls are discussed later, in Section 3.2. The key information
elements for a CRAVA inversion run is:
Seismic data.
Wavelet.
Signal/noise ratio.
Inversion volume.
Background model.
Correlation structures.
Since CRAVA is designed to estimate any information that is not given, well data must also com-
monly be included.
3.1.1 Survey information
All information regarding the seismic data is gathered under the <survey> tag. This includes
the le names for seismic data les, wavelet information and signal-to-noise ratio for each angle
gather. As an example, it may look like this:
<crava>
<survey>
<segy-start-time> 2500.0 </segy-start-time>
<angle-gather>
<offset-angle> 16.0 </offset-angle>
<seismic-data>
<file-name> seismic/Cube16.segy </file-name>
</seismic-data>
</angle-gather>
<angle-gather>
<offset-angle> 28.0 </offset-angle>
<seismic-data>
<file-name> seismic/Cube28.segy </file-name>
CRAVA User Manual version 1.2 31
</seismic-data>
</angle-gather>
</survey>
</crava>
The seismic data can be given on SegY-format, with a common offset time specied by the key-
word <segy-start-time> if the offset is different from 0. The rst value is used to represent the
interval from start-time to start-time + time-step, so with a start-time of 100ms, and 4ms sam-
pling, the rst value is used in the grid cell covering the interval 100-104ms. If we use seismic
data of another format than SegY, the <segy-start-time> command is not used. The le format
is detected automatically by CRAVA.
For each available angle, the rest of the information is gathered under an <angle-gather> tag,
one for each offset. The actual angle is given by <offset-angle>.
3.1.1.1 Seismic data
The name of the seismic data le is given with <file-name>, as seen in Section 3.1.1. Naturally,
seismic data is always required when running an inversion. By default, CRAVA recognises four
SegY formats; Seisworks, Charisma, SIP and IESX, see Table 3.1.
Table 3.1. SegY formats recognised by Crava
Name X Y IL XL CoordScal CoordSys
SeisWorks 73 77 9 21 71 UTM
Charisma 73 77 5 21 71 UTM
IESX 73 77 221 21 71 UTM
SIP 181 185 189 193 71 UTM
You are also allowed to dene your own format using the <segy-format> command. A standard
format is given by <standard-format>. Possible arguments are seisworks, iesx, charisma or
SIP. Modications to the chosen standard format can be given by the following commands:
<location-x>, <location-y>, <location-il>, <location-xl> and <bypass-coordinate-scaling>.
For more information on how to use this, see <segy-format> in the reference manual chapter.
Other le formats recognised by CRAVA are storm, Sgri and crava.
3.1.1.2 Wavelet
To invert the seismic data, we need a wavelet for each angle. This wavelet can be read from le,
using the <wavelet> and <file-name> commands like this:
<angle-gather>
<offset-angle> 16.0 </offset-angle>
<seismic-data>
<file-name> seismic/Cube16.segy </file-name>
</seismic-data>
<wavelet>
<file-name> wavelets/wavelet16.txt </file-name>
</wavelet>
</angle-gather>
We can read wavelets on JASON and NORSAR format.
CRAVA User Manual version 1.2 32
The Ricker wavelet is implemented in CRAVA, and can be used by the command <ricker>. The
peak frequency is given as argument.
If the <wavelet> command is not given, or given without <file-name>, the wavelet is estimated.
See Section 2.3 for how this is done. If the wavelet is given on le, but not scaled, the command
<scale> should be used if the scale is known, otherwise, the scale can be estimated by using the
<estimate-scale> command. If none of these are specied, the wavelet will be used as it is on
le.
3.1.1.3 Signal/noise ratio
This ratio is given with <signal-to-noise-ratio>. If this command is not given, the ratio is
estimated. Note that we dene the signal to noise ratio as the data variance divided by the error
variance, where the data variance is model variance plus error variance.
3.1.2 Inversion volume
The volume used for inversion is given horizontally by a rectangle, and vertically bounded by a
top and base surface. It is dened by the command <output-volume> under <project-settings>.
Typically, it may look something like this:
<crava>
<project-settings>
<output-volume>
<utm-coordinates>
<reference-point-x> 403050.0 </reference-point-x>
<reference-point-y> 7211900.0 </reference-point-y>
<length-x> 500.0 </length-x>
<length-y> 500.0 </length-y>
<angle> 23.627 </angle>
<sample-density-x> 50.0 </sample-density-x>
<sample-density-y> 50.0 </sample-density-y>
</utm-coordinates>
<interval-two-surfaces>
<top-surface>
<time-file> horizons/FlatTop_3100ms.storm </time-file>
</top-surface>
<base-surface>
<time-file> horizons/FlatBase_3600ms.storm </time-file>
</base-surface>
<number-of-layers> 125 </number-of-layers>
</interval-two-surfaces>
</output-volume>
</project-settings>
</crava>
3.1.2.1 Lateral extent
The lateral extent of the inversion volume is specied by the command <area-from-surface>,
<utm-coordinates>, or <inline-crossline-numbers>. The command <utm-coordinates> de-
scribes a rectangle, which may be rotated relative to the seismic data. It has the following param-
eters, which must all be specied:
CRAVA User Manual version 1.2 33
<reference-point-x> is the UTM x-coordinate of one corner of the area.
<reference-point-y> is the UTM y-coordinate of the same corner.
<length-x> is the extent of the area along the local x-axis.
<length-y> is the extent of the area along the local y-axis.
<angle> is the angle between the direction of the UTM x-axis and the local x-axis. Positive
angles are counterclockwise.
<sample-density-x> is the length of one grid cell along the local x-axis.
<sample-density-y> is the length of one grid cell along the local y-axis.
If the grid has the same rotation as the SegY volume read as input, output SegY volumes will have
correct in-lines and cross-lines. Otherwise, these numbers are just counting fromthe initial corner.
The command <area-from-surface> contains only one parameter, <file-name>, the name of a
stormsurface le dening the lateral extent of the inversion volume. The last way to dene the in-
version area is by the command <inline-crossline-numbers>. By this command, the following
parameters can be used:
<il-start> is the starting inline number.
<il-end> is the ending inline number.
<xl-start> is the starting crossline number.
<xl-end> is the ending crossline number.
<il-step> is the inline interval.
<xl-step> is the crossline interval.
The parameters <il-start> and <xl-start> must be set if this command is used, the other pa-
rameters are optional. If they are not given, the numbers are taken from the Segy le containing
the rst seismic cube.
The area commands may be skipped altogether. The area will then be taken from the rst input
seismic data le, and dened as the smallest rectangle that covers all traces.
3.1.2.2 Top and base surfaces
The vertical extent is normally given by a top and a base surface in time, under the command
<interval-two-surfaces> , as shown in Section 3.1.2. The le name for the top surface is given
under command <top-surface>, <time-file>, and the base surface is given similarly under
<base-surface> , <time-file>. The le format is binary storm or ASCII Irap. Top and base time
values for the inversion interval can be given as constants instead of les, by using the command
<time-value>.
These surfaces also dene the default lateral correlation direction for the elastic parameters, with
the correlation being parallel to the top surface at the top, and base surface at the base. Between
this, we create a top- and base-conform grid, so that the number of grid cells in each trace is
constant, although the interval thickness may vary. This is shown in part B of Figure 3.1. The
inversion may be unstable if the resolution varies too much in different traces, so we recommend
that no trace interval is larger than twice the shortest interval. The number of layers between top
and base is given by the command <number-of-layers> .
CRAVA User Manual version 1.2 34
Figure 3.1. The layer structure of a (A) parallel top and base, (B) top- and base-conform compaction grid (C)
Uniform correlation structure in a cut grid (D) Compactional correlation structure in a cut grid.
By specifying a correlation surface, the correlation direction can be independent of the interval
surfaces, see part C of Figure 3.1 and Section 3.1.3.2. If this is done, there are no restrictions on the
differences in interval thickness.
The more exible approach where a compactional correlation structure is specied independent
of the interval of interest (see part D of Figure 3.1) is currently not implemented.
If only one surface is known, the command <interval-one-surface> can be used to invert an
interval with top and base parallel to this surface. See Section 4.2.1.2 in the reference guide for
more details. Note that depth conversion and correlation surfaces will not be available in this
mode, so the lateral correlation will be parallel to this surface, as illustrated in part Aof Figure 3.1.
3.1.2.3 Depth conversion
The <output-volume> command is also where the depth conversion is specied. Additional in-
formation for depth conversion has only to be given under <interval-two-surfaces>, since the
lateral area is the same. To do a depth conversion, one of the following must be given:
Reference surface in depth (either top or base), and a velocity cube.
Both top and base surface in depth. In this case, we assume constant velocity along each trace,
computed from the time and depth surfaces.
Both top and base surface in depth, and a velocity cube. In this case, we use the cube for
relative velocity in a trace, and scale it to match the interval length.
Reference surfaces in depth are given with the tags <depth-file> under <top-surface> and/or
<base-surface>. The velocity cube can be read from le with the command <velocity-field>.
CRAVA User Manual version 1.2 35
Alternatively, the command <velocity-field-from-inversion> can be used to specify that V
p
frominversion should be used for velocity. With depth conversion, the <interval-two-surfaces>
command may look like this:
<interval-two-surfaces>
<top-surface>
<time-file> FlatTop_3100ms.storm </time-file>
<depth-file> FlatTop_3100ms.storm </depth-file>
</top-surface>
<base-surface>
<time-file> FlatBase_3600ms.storm </time-file>
<depth-file> FlatBase_3800ms.storm </depth-file>
</base-surface>
<velocity-field> velocity.storm </velocity-field>
<number-of-layers> 125 </number-of-layers>
</interval-two-surfaces>
3.1.3 Prior model
Since seismic data only contain information about relative elastic parameters, the absolute level
needs to be set with a background model. In a Bayesian inversion setting, the background model
is the prior expectation. We also need the prior covariance, which is given by the covariance of the
parameters, the lateral correlation and the temporal correlation, as described in Section 1.4. In the
model le, all this is gathered under the <prior-model> command, which may look something
like this:
<crava>
<prior-model>
<background>
<vp-file> input/background/CravaBgVp.storm </vp-file>
<vs-file> input/background/CravaBgVs.storm </vs-file>
<density-file> input/background/CravaBgRho.storm </density-file>
</background>
<lateral-correlation>
<variogram-type> genexp </variogram-type>
<power> 1 </power>
<angle> 0 </angle>
<range> 2500 </range>
<subrange> 2500 </subrange>
</lateral-correlation>
</prior-model>
</crava>
3.1.3.1 Background model
The background model is given under the <background> command. It can be given from le,
using <vp-file>, <vs-file> and <density-file>. These les should either be on Storm, crava,
Sgri or SegY format. Alternatively, constant values can be used for background model, specied
with <vp-constant>, <vs-constant> and <density-constant>. Any combinations of les and
constants are also accepted. If none of these are given, the background model will be estimated.
CRAVA User Manual version 1.2 36
3.1.3.2 Covariances
As shown in Section 1.4 the prior covariance structure for the elastic parameters consists of three
parts:
1. A 3x3 covariance matrix for point-wise covariance between the parameters. May be read
from ASCII le using the command <parameter-correlation>.
2. A temporal correlation vector, length equal to number of layers in grid, n
t
. May be read from
ASCII le using the command <temporal-correlation>.
3. A lateral correlation structure. May be given as a parametric variogram using the command
<lateral-correlation>.
By default, the two rst are estimated from well data, and the lateral correlation structure is set to
a isotropic exponential variogram with range 1000. The reason for the latter choice is that this is
hard to estimate, see Section 2.2.2 for details. The most common to override is the lateral correla-
tion, where the variogram used for petro-physical modelling is a good choice.
3.1.4 Well data
Unless all information about wavelet, signal to noise and correlations are specied, well data are
needed for estimation. Wells are given with the command <well-data>, and may look like this:
<crava>
<well-data>
<log-names>
<time> TWT </time>
<dt> DT </dt>
<dts> DTS </dts>
<density> RHOB </density>
</log-names>
<well>
<file-name> input/logs/ed6406_3-3_cut.rms </file-name>
</well>
<well>
<file-name> input/logs/ed6506_12-3_cut.rms </file-name>
</well>
<well>
<file-name> input/logs/ed6506_12-8_cut.rms </file-name>
</well>
<well>
<file-name> input/logs/ed6506_12-S4H_cut.rms </file-name>
</well>
</well-data>
</crava>
There are two main elements here. The rst is a well log interpretation, given by <log-names>,
which tells CRAVA which headers to look for. The following logs are needed:
Two way time log, specied with <time>.
V
p
log, either given by <vp> or <dt>. The latter is used for DT-logs.
V
s
log, either given by <vs> or <dts>. The latter is used for DTS-logs.
CRAVA User Manual version 1.2 37
Density log, given by <density>.
In addition, if facies probabilities are computed, a facies log is needed. This is specied with the
tag <facies>.
The wells are given with the command <file-name> under <well>, which is given once for each
well. The reason for this is that additional information may be given for each well. Well les
should be on NORSAR or RMS-format.
Each well may be moved to its optimal location using <optimize-position> under <well>, tak-
ing the arguments <angle> and <weight> which allow the user to assign different weights to the
different angle gathers for each well. The maximum allowed offset and vertical shift for moving
wells is specied in <maximum-offset> and <maximum-shift> under <well-data>, with default
values of 250 m and 11 ms, respectively.
The command <synthetic-vs-log> tells whether the V
s
log is synthetic. If not specied, it will
be detected from rank correlation with V
p
. The command <filter-elastic-logs> is used to do
multi-parameter-ltering of the elastic logs in this well after the inversion.
3.1.5 I/O settings
Two commands are used to specify the directory for input les. <top-directory> gives the work-
ing directory for the model le. The command <input-directory> is used to specify directory
name for root directory for input les. The name is given relative to <top-directory>.
Under <advanced-settings>, the command <use-intermediate-disk-storage> can be used to
limit the memory usage when running large CRAVA jobs. Abuilt-in smart-swap is then activated.
The option has largest effect on Microsoft Windows.
Under <project-settings>, <io-settings> and <other-output>, the command <error-file>
writes all errors to a separate le, in addition to the log le. The command <task-file> writes all
tasks to a separate le, in addition to the log le.
3.1.6 Output
Output is controlled under <io-settings> under <project-settings>. Here, you may set the
output directory using <output-directory>, and you may also specify a prex for all output
les using <file-output-prefix>.
Except for the log le which is placed directly under the output directory, all les output by crava
are placed in sub-directories. These sub-directories are
output-directory / wells
/ background
/ wavelets
/ seismic
/ velocity
/ correlations
/ inversionresults
There are two main output formats: Grid output, controlled by <grid-output>, and well output
controlled by <well-output>. The output section may look something like this:
<crava>
<project-settings>
<io-settings>
<file-output-prefix> CRAVA_ </file-output-prefix>
CRAVA User Manual version 1.2 38
<grid-output>
<format>
<segy> yes </segy>
</format>
<domain>
<time> yes </time>
<depth> yes </depth>
</domain>
<elastic-parameters>
<vp> yes </vp>
<vs> yes </vs>
<density> yes </density>
<background> yes </background>
</elastic-parameters>
</grid-output>
<well-output>
<wells> yes </wells>
<blocked-wells> yes </blocked-wells>
</well-output>
</io-settings>
</project-settings>
</crava>
3.1.6.1 Grid output
Different elastic parameters can be given as grid output. In addition, the estimated background
model may be written as grids. This is controlled by the <elastic-parameters> command under
<grid-output>. See Section 4.2.3.4.3 for a full list of possible grids. If this command is not used,
V
p
, V
s
and density will be written. Output of original and synthetic seismic data can be given
by the <seismic-data> command. Other grids can be requested for output by the command
<other-parameters>, for example correlations.
The grid format may be controlled using <format>. Here the yes/no parameters <storm>, <segy>,
<sgri>, <crava> and <ascii> can be used to decide if grids should be written on storm- (RMS),
segy-, Sgri-, crava- or ASCII-format. You may choose several formats for one run; all grids will
be written on all selected formats. The Segy format can be controlled by the <segy-format> com-
mand, in the same way as for input data, described in Section 3.1.1.1. Note that correlation grids
make sense only in storm format. Default output format is storm. The crava format is a binary
format only to be used with CRAVA. It can be read and written from CRAVA, and is useful if the
output from a CRAVA run should be used as input to another CRAVA run because the format is
fast to read.
By using the <domain> option, output may be written in time domain <time> or depth domain
<depth> (requires parameters set under <output-volume>, see Section 3.1.2.3), or both. Again,
correlation grids only make sense in time domain, which is default.
3.1.6.2 Well output
Some versions of ltered elastic parameters (V
p
, V
s
and density) in wells can be generated by the
<well-output> command. The wells can be given in two different formats, RMS or NORSAR.
This is controlled by the <format> command. The logs written are
Raw elastic logs.
CRAVA User Manual version 1.2 39
Elastic logs ltered to background frequency.
Elastic logs ltered to seismic frequency.
Elastic logs ltered with facies prediction lter (if available).
Facies log (if available).
The wells can either be written with original sampling density, using <wells>, or matching the
internal grid resolution, using <blocked-wells>.
3.1.7 Actions
The nal information that is needed for a CRAVA run is what the run is supposed to do. This is
controlled by <actions>. The <mode> keyword denes the purpose of this run and should be set to
"inversion" when doing inversion. Other options are "estimation", see Section 3.3 and "forward",
see Section 3.6. When inversion is chosen, <inversion-settings> can be used to control basic
aspects of the inversion. It may look something like this:
<crava>
<actions>
<mode> inversion </mode>
<inversion-settings>
<prediction> yes </prediction>
<simulation>
<seed> 150570 </seed>
<number-of-simulations> 10 </number-of-simulations>
</simulation>
</inversion-settings>
</actions>
</crava>
The command <prediction> can be used to turn predictions on or off. By default, the prediction
will be generated. A number of full frequency stochastic realisations of the inversion can be ob-
tained by specifying <number-of-simulations> under <simulation>. The seed for the random
generator can also be given here, with the <seed> command. Changing the seed will give a new
set of realisations.
The command <kriging-to-wells> can be used to krige realisations to well data. By default, this
is done if the <simulation> command is used.
3.1.8 Standard grid formats
3D-grid: Crava can read and write the following 3D grid formats: SegY, storm, Sgri (NOR-
SAR) and crava (internal) . On output, the format is controlled by a <format> keyword. On
input, the program automatically detects the format, although it may need help to correctly
read SegY formats, using the <segy-format> keyword. The default output format is storm.
Surfaces: The standard format for reading and writing surfaces is binary storm, but CRAVA
can also read Roxar ASCII surfaces (formerly known as ASCII Irap classic). In addition some
surfaces related to 3D-inversion are read as sgri (NORSAR). On input, the surface format will
be automatically detected.
Wavelets: Wavelets are either on JASON or NORSAR format, both for reading and writing.
Auto-detect is used on reading, use <format> for writing. The default output format is JA-
SON.
CRAVA User Manual version 1.2 40
Wells: Wells are either on RMS or NORSARformat, both for reading and writing. Auto-detect
is used on reading use <format> to control writing. The default output format is RMS.
3.2 Advanced inversion options
Although CRAVA is mainly intended as a simple and fast inversion tool, there are still some
options to control the inversion, and to access more sophisticated approaches. Most of these are
covered here, see also Section 4.2.4 for details about <advanced-settings>.
3.2.1 Non-stationary wavelet and noise
Although the FFT-algorithm which is at the core of CRAVA requires stationarity, this does not
mean that the entire inversion has to be stationary, as discussed in Section 2.6. We allow lateral
variations in wavelet amplitude, wavelet shift and signal to noise ratio.
The local wavelet transformations t well within the core framework of CRAVA, but the local
noise requires some approximations, as explained in section Section 2.6.2. This means that local
noise only improves the nal result if the local variations are substantial. We do not recommend
using local noise if the variation in noise level is less than 15%.
Unlike the basic level, where parameters that were not specied automatically got estimated, the
use of local wavelets or local noise must be explicitly triggered. For wavelets, this is done with
the <local-wavelet> command under <wavelet>. There are four alternatives: <local-wavelet>:
1. <shift-file> gives a le name for a map giving the local shifts.
2. <estimate-shift> will estimate a local shift map when set to "yes".
3. <scale-file> gives a le name for a map giving the local scale.
4. <estimate-scale> will estimate a local scale map when set to "yes".
Naturally, the shift can not be both given and estimated, the same holds for scale. Note that you
may choose to use only shifts, omitting both scale keywords, or use only scale.
The local noise is triggered similarly, by using one of two commands under <angle-gather>:
1. <local-noise-scaled> gives a le name for a map with the local scaling of the signal to
noise ratio.
2. <estimate-local-noise> estimates the local scaling of the signal to noise ratio if set to "yes".
3.2.2 PS-seismic and reection approximations
By default, CRAVA assumes that the input seismic data are PP, but PS data can also be used.
For both cases, we use the linearised Aki-Richards approximation, see Equation 1.1. The type
of seismic data is indicated by using the <type> command under<seismic-data>. Here, <type>
should be either "pp" or "ps". Note that PS data must also be aligned in PP-travel time, as no such
alignment is done internally by CRAVA.
Instead of using the default reection approximation, the user may supply the parameters to
compute the reections. We always assume that for a given angle and seismic type, the reections
can be computed from the equation
c(x, t, ) = a
Vp
()

t
lnV
p
(x, t)
+ a
Vs
(x, t, )

t
lnV
s
(x, t)
+ a

(x, t, )

t
ln(x, t).
(3.1)
CRAVA User Manual version 1.2 41
Table 3.2. Default intervals for valid well log values.
Min Max
V
p
1300 7000
V
s
200 4200
1.4 3.3
Table 3.3. Default intervals for valid well log variances.
Min Max
Var(ln(V
p
)) 5 10
4
250 10
4
Var(ln(V
s
)) 10 10
4
500 10
4
Var(ln()) 2 10
4
100 10
4
The coefcients a
Vp
, a
Vs
and a

can be read from le, using the <reflection-matrix> command

under <advanced-settings>. This le should have one line for each seismic data le, and each
line should have the three coefcients for one set of seismic input data. The order of the lines
should be the same as the order of the seismic data in the <survey> command.
3.2.3 Well quality checks
Well logs are often faulty, and CRAVA has some safety mechanisms to detect this. The primary
mechanism is to detect extreme values, and set these undened. The default upper and lower
bounds accepted in well logs are shown in Table 3.2. These can be overridden using the com-
mand <allowed-parameter-values> under <well-data>. Here, <minimum-vp>, <minimum-vs>,
<minimum-density> and the corresponding maximum values can be given.
Some logs may stay within reasonable values, but have too much or too little variation, also in-
dicating that something is wrong. We therefore calculate the variance of the logarithm of the log
minus the background, and if this is outside reasonable bounds, an error is triggered. The de-
fault bounds shown in Table 3.3 can be overridden with <allowed-parameter-values>, using
<minimum-variance-vp>, and so forth.
3.2.4 Generate synthetic seismic from inversion data
It is possible to generate synthetic seismic by a forward modelling with the V
p
, V
s
and density
resulting from the inversion. This is done by the command <synthetic> under <seismic-data>
under <grid-output>, <io-settings>, in command <project-settings>.
3.3 Estimation
As mentioned, CRAVA can estimate many of the needed parameters. There are several commands
that control the estimation behaviour for wavelets, noise and background model. Note that the
correlations will always be estimated as explained in Section 2.2.2 from all available well logs,
and do not have any further controls.
3.3.1 Estimation mode
If you only want to do the estimation, in order to check the quality of the estimates, you can
use "estimate" in the <mode> command. Using this, CRAVA will perform the initial model build-
ing tasks and estimate needed information, but terminate once all information needed for inver-
sion is estimated. When running in estimation mode, you can also control the main estimation
aspects using the <estimation-settings> command. This allows you to control which of the
main estimation tasks should be carried out, setting yes or no for <estimate-correlations>,
CRAVA User Manual version 1.2 42
<estimate-wavelet-or-noise> or <estimate-background>. Parameters with a "no" will not be
estimated unless needed for other estimations. Note that in estimation mode, all estimated pa-
rameters are written to le, regardless of output settings.
3.3.2 Wavelet and noise estimation
Wavelets and noise are estimated together. These parameters will only be estimated from wells
that are vertical or close to vertical, since this allows comparing synthetic seismic from well logs
with one or a couple of traces, and thus reduces alignment issues. The angle limit can be con-
trolled by the <maximum-deviation-angle> command under <well-data>. In addition, wells can
be excluded individually, by setting <use-for-wavelet-estimation> to "no" under <well>.
By default, the wavelet and noise are estimated in the region between the top and base surface for
the inversion area. This may be limited using the <wavelet-estimation-interval> command
under <survey>, where a separate set of restricting surfaces are given with <top-surface-file>
and <base-surface-file>. Since these are under <wavelet>, they can be different for each angle.
Some wavelet estimation options are also found under the command <advanced-settings> given
under <project-settings>. These are
<wavelet-tapering-length> which controls the length of the wavelet (in ms).
<minimum-relative-wavelet-amplitude> which nds the wavelet length, by setting the cut-
off size for edge peaks relative to centre peak.
<maximum-wavelet-shift> which controls how far the wavelet can be shifted locally, see
Section 2.3.
Prior information for local wavelet modelling is given by the <local-wavelet> command under
<prior-model>. Here, lateral correlation is specied as a 2D variogram for the lateral correlation
in local wavelet modelling by the command <lateral-correlation>.
Output of wavelets is controlled by the <wavelet-output> command under <io-settings>. Two
different formats can be specied, JASON and NORSAR. Estimated wavelets for each well is
written by the <well-wavelets> command. The command <global-wavelets> writes global
wavelets for each seismic angle. If local wavelet is requested, the command <local-wavelets>
writes estimated local wavelet shift and scale surfaces.
Output of estimated local noise surface is given by the command <local-noise> given under
<io-settings> and <other-output>.
3.3.3 Background model estimation
The background model is estimated as a low frequency vertical trend. The trend is given in rela-
tive depth in the inversion volume, so the trend value along the top and base surface is constant.
At well locations, the trend is kriged to the well logs. By default, all wells are used for background
model estimation, but can be excluded by specifying "no" for <use-for-background-trend> un-
der <well> for individual wells.
The background estimation can be controlled by some commands under <background>. These
are:
<velocity-field> takes an external velocity eld, typically from migration, and uses as V
p
background, kriged to low frequency V
p
well logs.
<lateral-correlation> gives a variogram for the kriging. Larger ranges extends the inu-
ence of well data further away from wells.
CRAVA User Manual version 1.2 43
<high-cut> allows specication of a maximum frequency for the background model.
3.3.4 Prior correlations
Prior correlations between V
p
, V
s
and density, and autocorrelation are estimated. they can be writ-
ten to le by the command <prior-correlations> under <project-settings>, <io-settings>
and <other-output>.
3.4 3D wavelet
Instead of the traditional 1D wavelet a 3D wavelet can be used in CRAVA. This wavelet is con-
structed by a 1D pulse and a 3D wavelet lter based on illumination vectors. The use of a 3D
wavelet is invoked by the keyword <wavelet-3d>.
The 1D pulse used to construct the 3D wavelet, can be specied by the <file-name> keyword.
If a le is not specied, this pulse is estimated. Either way, a lter le must be given by the
keyword <processing-factor-file-name>. This lter contains the amplitude effects related to
the illumination vectors and is given on Sgri format. One or two lters are given. The rst, and
necessary, contains frequency-independent amplitude effects, while the second, if given, contains
frequency-dependent amplitude effects.
For noise estimation in the 3D wavelet setting, a propagation lter le is given by the keyword
<propagation-factor-file-name>. At present, this is not used. The keyword <stretch-factor>
can be given, but is not used since the stretch factor is automatically calculated from the offset
angle.
When estimating the 1D pulse, seismic data from a neighbourhood of the well can be used. The
distance from the well, specied in meters, that data are collected is given by the commands
<estimation-range-x-direction> and <estimation-range-y-direction>. The default is that
only data from the well position are used.
When 3D wavelet is used, certain parameters related to the mapping between time and depth
are needed. This occurs since the illumination vectors relates to depth, while CRAVA works in
time. These settings are given under keyword <time-to-depth-mapping-for-3d-wavelet> in
<project-settings>. A reference depth for the target area must be specied under the key-
word <reference-depth>. This depth represents a constant surface which in time normally will
refer to a variable surface due to varying velocity above this depth. This reference surface is
given by <reference-time-surface> and can be given on Storm- or Sgri-format. The parame-
ter <average-velocity> refers to the average velocity in the target area which is needed for the
estimation of the pulse.
3.5 Facies prediction
An important feature in CRAVA is the ability to create facies probabilities. This requires that
<mode> is set to "inversion", and is triggered by the <facies-probabilities> command under
<inversion-settings>.
The facies probabilities are computed based on the inversion results and a distribution for in-
version values given facies computed from ltered well logs, where the lter is dened by the
inversion. See Section 1.4.1. Probability volumes will be computed for all facies seen in wells, and
if the command <facies-probabilities-with-undef> is used, an additional undened proba-
bility cube is also generated, indicating areas where the inversion values are too far away from
well data to make reliable predictions.
Except from the trigger in <inversion-settings>, all parameters related to facies probabilities
CRAVA User Manual version 1.2 44
are given under <facies-probabilities> under <prior-model>. Here is an example:
<crava>
<prior-model>
<facies-probabilities>
<use-vs> yes </use-vs>
<use-prediction> yes </use-prediction>
<use-absolute-elastic-parameters> yes </use-absolute-elastic-parameters>
<estimation-interval>
<top-surface-file> input/horizons/facies_top.storm </top-surface-file>
<base-surface-file> input/horizons/facies_base.storm </base-surface-file>
</estimation-interval>
<prior-probabilities>
<facies>
<name> sand </name>
<probability> 0.4 </probability>
</facies>
<facies>
<name> shale </name>
<probability> 0.6 </probability>
</facies>
</prior-probabilities>
</facies-probabilities>
</prior-model>
</crava>
Figure 3.2. A sample facies probability estimation. The black curve gives the probability of sand in a well
containing sand (orange), shale (green) and crevasse (brown). In the left gure the well was not included in
the facies probability estimate whereas it was included in the right gure.
CRAVA User Manual version 1.2 45
If <use-absolute-elastic-parameters> is set to yes, facies probability is computed based on
inverted parameters including background model. If the distribution for elastic parameters for
each facies is constant over the inversion volume, using absolute values is more stable. However,
if there are trends in the elastic parameters, the relative values are more robust. The default is to
use relative values.
The interval to use for facies probability calculations can be set with <estimation-interval>,
similar to estimation interval for wavelets. Parallel to the wavelet case, wells may also be excluded
using the <use-for-facies-probabilities> command under <well>.
3.5.1 Prior probabilities
In order to get reliable probabilities, we need good prior probabilities. By default, CRAVA com-
putes the average fraction of each facies in the relevant interval of the wells. This can be over-
ridden using the <prior-probabilities> command, which allows specication of these. Note
that probabilities must be given for each facies. Probabilities can either be given globally, with
<probability>, or as a full 3D trend, using <probability-cube>. The latter takes a grid le as
argument, and the corresponding grid must cover the inversion volume. A prior value for unde-
ned facies is set with the command <uncertainty-level>.
3.5.2 Output parameters
Facies probabilities can be written to le by using the command <facies-probabilities> or
<facies-probabilities-with-undef> given under <project-settings>, <io-settings>, <grid-output>
and <other-parameters>. <facies-probabilities> gives probabilities for the existing facies
that sum up to one. This is recommended for use in RMS. <facies-probabilities-with-undef>
includes probability for undened facies, and this cube will be the most correct one. <facies-likelihood>
writes the likelihood for inverted seismic for each facies, p(m|f), while <seismic-quality-grid>
writes a grid kriged from values for t between facies probabilities and facies observed in each of
the wells.
Rock physics distributions can be written by the command <rock-physics-distributions> un-
der <io-settings> and <other-output>. The distributions are written per facies, in a STORM
grid with V
p
, V
s
and density as axes.
3.6 Forward modelling
A minor functionality in CRAVA is that it can do forward modelling, showing what seismic re-
sponse the programwould expect froma given set of elastic parameters. This is triggered by using
"forward" as <mode>. In this mode, we generate synthetic seismic data fromthe given background
volumes. A le for forward modelling looks like this:
<crava>
<actions>
<mode> forward </mode>
</actions>
<survey>
<angle-gather>
<offset-angle> 0 </offset-angle>
<wavelet>
<file-name> wavelets/ricker.txt </file-name>
</wavelet>
CRAVA User Manual version 1.2 46
</angle-gather>
<angle-gather>
<offset-angle> 10 </offset-angle>
<wavelet>
<file-name> wavelets/rickershift.txt </file-name>
</wavelet>
</angle-gather>
</survey>
<prior-model>
<earth-model>
<vp-file> background/Vp.storm </vp-file>
<vs-file> background/Vs.storm </vs-file>
<density-file> background/Rho.storm </density-file>
</earth-model>
</prior-model>
<project-settings>
<output-volume>
<!-- Lateral dimensions are extracted from the volume for Vp -->
<interval-two-surfaces>
<top-surface>
<time-file> horizons/top.irap </time-file>
</top-surface>
<base-surface>
<time-file> horizons/base.irap </time-file>
</base-surface>
<number-of-layers> 250 </number-of-layers>
</interval-two-surfaces>
</output-volume>
<io-settings>
<grid-output>
<format>
<segy> yes </segy>
</format>
</grid-output>
</io-settings>
</project-settings>
</crava>
Note that no seismic les are given under <survey>, as these are now computed. No wells are
used, since nothing can be estimated here. We need the angles to generate seismic data for, the
corresponding wavelets, the elastic parameters (given as earth model), and the volume. Instead
of using one of the commands dening volume, the volume can be taken from V
p
. The output
format can be controlled using <io-settings>, as can input- and output-directory. Other input
will be ignored.
CRAVA User Manual version 1.2 47
4 Model le reference manual
The numbering shows the command grouping. A command with no sub-numbering expects a
value to be given, otherwise, it is only a grouping of other commands.
File names are currently given with a path relative to the directory settings in <project-settings>-
<io-settings>-<input/output/top-directory>. If these are not given, the path will always be rela-
tive to the working directory.
All commands are optional, unless otherwise stated. A necessary command under an optional is
only necessary if the optional is given.
Standard grid formats for surfaces and 3D grids are given in Section 3.1.8.
4.1 <actions> (necessary)
Description: Controls the main purpose of the run.
Argument: Elements specifying the main purpose
4.1.1 <mode> (necessary)
Description: Inversion: Invert seismic input data to elastic parameters and/or facies probabilities.
Needs seismic data and volume, all other missing data will be estimated. Forward: Create
seismic response from background model. Not able to estimate anything. Estimation: Checks
input data and performs estimation of lacking information for inversion, but stops before
inversion.
Argument: inversion, forward or estimation
4.1.2 <inversion-settings>
Description: Controls aspects of the inversion. Only valid with the <mode> inversion above.
Argument: Elements controlling the inversion
Default:
4.1.2.1 <prediction>
Description: Controls whether predicted elastic parameters will be generated.
Argument: yes or no
Default:
4.1.2.2 <simulation>
Description: Controls aspects of the simulation of elastic parameters.
Argument: Elements controlling the simulation of elastic parameters
4.1.2.2.1 <seed>
Description: A number used to initialise the random generator. Running a model le with a given
seed will give the same simulation results each time.
Argument: Integer
Default: 0
CRAVA User Manual version 1.2 48
4.1.2.2.2 <seed-file>
Description: Alternative to <seed>. This is an ASCII le containing a number. At the termination
of the run, the le will be overwritten with a seed generated by the random generator. Thus,
a model le using this will generate different simulation results on sequential runs.
Argument: ASCII le containing a number
Default:
4.1.2.2.3 <number-of-simulations>
Description: Integer value giving the number of stochastic realizations to generate.
Argument: Integer
Default: 0
4.1.2.3 <kriging-to-wells>
Description: Should the realizations be kriged to well data?
Argument: yes or no
Default: yes if not the <simulation> command is used.
4.1.2.4 <facies-probabilities>
Description: Should facies probabilities be estimated?
Argument: yes or no
Default: no
4.1.3 <estimation-settings>
Description: Controls what will be estimated. Only valid with the <mode> estimation. Note that
these commands can only turn off estimations - a parameter that is given will not be esti-
mated even if it says so here.
Argument: Elements controlling what to estimate
Default:
4.1.3.1 <estimate-background>
Description: If no, background will not be estimated unless needed for other estimation.
Argument: yes or no
Default: yes
4.1.3.2 <estimate-correlations>
Description: If no, correlations will not be estimated unless needed for other estimation.
Argument: yes or no
Default: yes
4.1.3.3 <estimate-wavelet-or-noise>
Description: If no, wavelets and/or noise will not be estimated unless needed for other estima-
tion.
Argument: yes or no
Default: yes
4.2 <project-settings> (necessary)
Description: Controls inversion volume, output and advanced program settings.
Argument: Elements controlling inversion volume, output and advanced program settings
CRAVA User Manual version 1.2 49
Default:
4.2.1 <output-volume> (necessary)
Description: Denes the core inversion volume. All grid output will be given in this volume.
Argument: Elements dening the core inversion volume.
Default:
4.2.1.1 <interval-two-surfaces>
Description: One way to give the top and bottom limitations. Must be used if output in depth
domain is desired. This or <interval-one-surface> must be given.
Argument:
Default:
4.2.1.1.1 <top-surface> (necessary)
Description: File name(s) for top surface le(s).
Argument: Elements controlling the top surface
Default:
4.2.1.1.1.1 <time-file>
Description: File name for standard surface le giving top surface in time. This or <time-value>
must be given.
Argument: File name
Default:
4.2.1.1.1.2 <time-value>
Description: Value giving the top time for the inversion interval. This or <time-file> must be
given.
Argument: Value
Default:
4.2.1.1.1.3 <depth-file>
Description: File name for standard surface le giving top surface in depth.
Argument: File name
Default:
4.2.1.1.2 <base-surface> (necessary)
Description: File name(s) for base surface le(s).
Argument: Elements controlling the base surface
Default:
4.2.1.1.2.1 <time-file>
Description: File name for standard surface le giving base surface in time. This or <time-value>
must be given.
Argument: File name
Default:
CRAVA User Manual version 1.2 50
4.2.1.1.2.2 <time-value>
Description: Value giving the base time for the inversion interval. This or <time-file> must be
given.
Argument: Value
Default:
4.2.1.1.2.3 <depth-file>
Description: File name for standard surface le giving base surface in depth.
Argument: File name
Default:
4.2.1.1.3 <number-of-layers>
Description: Integer value giving how many layers to use between top and base surface.
Argument: Integer
Default:
4.2.1.1.4 <velocity-field>
Description: File name for standard 3D grid le. Gives more detailed depth conversion informa-
tion. Without this, constant velocity per trace is used. If only one depth surface is given, this
is used to compute the other. Otherwise, the depth interval will always match both surfaces,
but the velocity eld is scaled and used for internal depth computations. Can not be used
with <velocity-field-from-inversion>.
Argument: File name
Default:
4.2.1.1.5 <velocity-field-from-inversion>
Description: If given, velocity eld frominversion is used for depth conversion. See <velocity-field>
for details on how this is done. Can not be used with <velocity-field>.
Argument: yes or no
Default:
4.2.1.2 <interval-one-surface>
Description: Using this command gives parallel top and base of inversion interval. This or <interval-two-surfaces>
must be given.
Argument: Elements for parallel top and base inversion interval
Default:
4.2.1.2.1 <reference-surface>
Description: File name for standard surface le. The top and base surfaces for the inversion inter-
val will be parallel to this.
Argument: File name
Default:
4.2.1.2.2 <shift-to-interval-top>
CRAVA User Manual version 1.2 51
Description: Value giving the distance from reference surface to top surface. This value is added
to the reference surface to create the top surface.
Argument: Value
Default:
4.2.1.2.3 <thickness>
Description: Value giving the thickness of the inversion interval. This value is added to the top
surface to create the base surface.
Argument: Value
Default:
4.2.1.2.4 <sample-density>
Description: Value giving the thickness of a layer in the inversion interval. The thickness should
be divisible by this value.
Argument: Value
Default:
4.2.1.3 <area-from-surface>
Description: Inversion area can be dened by a surface. Then the name of the surface is given in
this command. Other ways to dene inversion area are by the commands <utm-coordinates>
or <inline-crossline-numbers>. If none of these commands are used, the area is dened by
the rst seismic data le, or from Vp if we do forward modelling.
Argument:
Default:
4.2.1.3.1 <file-name>
Description: File name for standard surface le.
Argument: File name
Default:
4.2.1.3.2 <snap-to-seismic-data>
Description: Find the smallest rectangular IL-XL box enclosing the entire surface and do the in-
version using these IL-XL values. This allows a user to specify an inversion area in UTM and
get the inversion volume aligned with seismic data.
Argument: yes or no
Default: no
4.2.1.4 <utm-coordinates>
Description: Describe area by UTM coordinates.
Argument:
Default:
4.2.1.4.1 <reference-point-x>
Description: Value giving the x-coordinate of a corner of the area.
Argument: Value
Default:
CRAVA User Manual version 1.2 52
4.2.1.4.2 <reference-point-y>
Description: Value giving the y-coordinate of a corner of the area.
Argument: Value
Default:
4.2.1.4.3 <length-x>
Description: Value giving the area length along the rotated x-axis.
Argument: Value
Default:
4.2.1.4.4 <length-y>
Description: Value giving the area length along the rotated y-axis.
Argument: Value
Default:
4.2.1.4.5 <sample-density-x>
Description: Cell size along the rotated x-axis.
Argument: Integer
Default:
4.2.1.4.6 <sample-density-y>
Description: Cell size along the rotated y-axis.
Argument: Integer
Default:
4.2.1.4.7 <angle>
Description: Orientation of the azimuth.
Argument:
Default:
4.2.1.4.8 <snap-to-seismic-data>
Description: Find the smallest rectangular IL-XL box enclosing the entire UTM specied area
and do the inversion using these IL-XL values. This allows a user to specify an inversion
area in UTM coordinates and get an inversion volume aligned with seismic data. Keywords
<sample-density-x> and <sample-density-y> are not needed when snapping is activated.
Argument: yes or no
Default: no
4.2.1.5 <inline-crossline-numbers>
Description: Describe area by inline and crossline numbers. il-start and xl-start must be given if
this command is used, the other variables are optional. The numbers which are not specied
are taken fromthe SegY le containing seismic data. The command is only working if seismic
data are given on SegY format.
Argument:
Default:
CRAVA User Manual version 1.2 53
4.2.1.5.1 <il-start>
Description: Start value for inline.
Argument:
Default:
4.2.1.5.2 <il-end>
Description: End value for inline.
Argument:
Default:
4.2.1.5.3 <xl-start>
Description: Start value for crossline.
Argument:
Default:
4.2.1.5.4 <xl-end>
Description: End value for crossline.
Argument:
Default:
4.2.1.5.5 <il-step>
Description: Step value for inline.
Argument:
Default:
4.2.1.5.6 <xl-step>
Description: Step value for crossline.
Argument:
Default:
4.2.2 <time-to-depth-mapping-for-3d-wavelet>
Description: Denes the mapping between pseudo-depth and local time in the target area for 3D
wavelet.
Argument: Reference depth, velocity and time surface for mapping
Default:
4.2.2.1 <reference-depth>
Description: Holds the z-value for the reference depth for target area
Argument: Depth in meter
Default:
4.2.2.2 <average-velocity>
Description: Holds the average velocity in the target area
Argument: Velocity in meter/second
Default:
CRAVA User Manual version 1.2 54
4.2.2.3 <reference-time-surface>
Description: File name for the time surface corresponding to the reference depth. Standard surface
format.
Argument: File name
Default:
4.2.3 <io-settings>
Description: Holds commands that deal with what output to give and where, and where to nd
input.
Argument: Elements controlling output and input
Default:
4.2.3.1 <top-directory>
Description: Directory name giving the working directory for the model le. Must end with direc-
tory separator.
Argument: Directory name
Default:
4.2.3.2 <input-directory>
Description: Directory name, relative to <top-directory>, for root directory for input les. Must
end with directory separator.
Argument: Directory name
Default:
4.2.3.3 <output-directory>
Description: Directory name, relative to <top-directory>, for root directory for output les. Must
end with directory separator.
Argument: Directory name
Default:
4.2.3.4 <grid-output>
Description: All commands related to output given as grids are gathered here.
Argument: Elements controlling output given as grids
Default:
4.2.3.4.1 <domain>
Description: Commands specifying which domain output should be in.
Argument: Elements controlling the output domain
Default:
4.2.3.4.1.1 <depth>
Description: Should output come in depth domain? Requires information under <interval-two-surfaces>.
Argument: yes or no
Default: no
4.2.3.4.1.2 <time>
Description: Should output come in time domain?
CRAVA User Manual version 1.2 55
Argument: yes or no
Default: yes
4.2.3.4.2 <format>
Description: Control of the format of output grids.
Argument: Elements controlling the format of output grids
Default:
4.2.3.4.2.1 <segy-format>
Description: Information about the segy format. By default CRAVA recognises SeisWorks, IESX,
SIP and Charisma, see Table 3.1.
Argument: Elements containing information about the segy format.
Default:
4.2.3.4.2.1.1 <standard-format>
Description: Giving the starting format for modications.
Argument: seisworks, iesx, charisma or SIP
Default: seisworks
4.2.3.4.2.1.2 <location-x>
Description: The byte location for the x-coordinate in the trace header.
Argument: Integer
Default:
4.2.3.4.2.1.3 <location-y>
Description: The byte location for the y-coordinate in the trace header.
Argument: Integer
Default:
4.2.3.4.2.1.4 <location-il>
Description: The byte location for the inline in the trace header.
Argument: Integer
Default:
4.2.3.4.2.1.5 <location-xl>
Description: The byte location for the crossline in the trace header.
Argument: Integer
Default:
4.2.3.4.2.1.6 <bypass-coordinate-scaling>
Description: Indicates whether coordinate scaling information should be used.
Argument: yes or no
Default:
CRAVA User Manual version 1.2 56
4.2.3.4.2.1.7 <location-scaling-coefficient>
Description:
Argument: Integer
Default:
4.2.3.4.2.2 <segy>
Description: Should grid output come as segy?
Argument: yes or no
Default:
4.2.3.4.2.3 <storm>
Description: Should grid output come as storm?
Argument: yes or no
Default: yes if <format> command is not given.
4.2.3.4.2.4 <crava>
Description: Should grid output come in crava binary format?
Argument: yes or no
Default:
4.2.3.4.2.5 <sgri>
Description: Should grid output come as storm sgri?
Argument: yes or no
Default:
4.2.3.4.2.6 <ascii>
Description: Should grid output come as storm ascii?
Argument: yes or no
Default:
4.2.3.4.3 <elastic-parameters>
Description: Controls which elastic grid parameters to output. All are yes or no.
Argument: Elements controlling which elastic parameters to output
Default: If this command is not given, vp, vs and density are written.
4.2.3.4.3.1 <vp>
Description:
Argument: yes or no
Default:
4.2.3.4.3.2 <vs>
Description:
Argument: yes or no
Default:
CRAVA User Manual version 1.2 57
4.2.3.4.3.3 <density>
Description:
Argument: yes or no
Default:
4.2.3.4.3.4 <lame-lambda>
Description:
Argument: yes or no
Default:
4.2.3.4.3.5 <lame-mu>
Description:
Argument: yes or no
Default:
4.2.3.4.3.6 <poisson-ratio>
Description:
Argument: yes or no
Default:
4.2.3.4.3.7 <ai>
Description:
Argument: yes or no
Default:
4.2.3.4.3.8 <si>
Description:
Argument: yes or no
Default:
4.2.3.4.3.9 <vp-vs-ratio>
Description:
Argument: yes or no
Default:
4.2.3.4.3.10 <murho>
Description:
Argument: yes or no
Default:
4.2.3.4.3.11 <lambdarho>
Description:
Argument: yes or no
Default:
CRAVA User Manual version 1.2 58
4.2.3.4.3.12 <background>
Description:
Argument: yes or no
Default:
4.2.3.4.3.13 <background-trend>
Description:
Argument: yes or no
Default:
4.2.3.4.4 <seismic-data>
Description: Controls which seismic data parameters to output. All are yes or no.
Argument: Elements controlling which seismic data to output
Default:
4.2.3.4.4.1 <original>
Description: Write original seismic data to le.
Argument: yes or no
Default:
4.2.3.4.4.2 <synthetic>
Description: Generate synthetic seismic from the inverted data.
Argument: yes or no
Default:
4.2.3.4.4.3 <residuals>
Description: Residuals computed in the actual inversion. Will not add up to original seismic data
when combined with synthetic seismic, due to some ltering and numerical imprecision.
Argument: yes or no
Default: no
4.2.3.4.4.4 <synthetic-residuals>
Description: Residuals computed by taking the original seismic and subtracting the synthetic seis-
mic. This means that these parameters are a matched set.
Argument: yes or no
Default: no
4.2.3.4.5 <other-parameters>
Description: Controls which other parameters to output. All are yes or no.
Argument: Elements controlling which other to output
Default:
4.2.3.4.5.1 <facies-probabilities>
Description: Write facies probabilities to le.
CRAVA User Manual version 1.2 59
Argument: yes or no
Default: yes if facies estimation is requested and <facies-probabilities-with-undef> is
not specied
4.2.3.4.5.2 <facies-probabilities-with-undef>
Description: Write facies probabilities with undened value to le.
Argument: yes or no
Default: no
4.2.3.4.5.3 <facies-likelihood>
Description: Write likelihood for inverted seismic for each facies, p(m|f).
Argument: yes or no
Default: no
4.2.3.4.5.4 <time-to-depth-velocity>
Description: Write time-to-depth velocity to le.
Argument: yes or no
Default:
4.2.3.4.5.5 <extra-grids>
Description: Temporary, will be replaced. Currently triggers writing of
Estimated background les in extended versions (go above and below inversion vol-
ume).
Estimated background les in standard volume.
Argument: yes or no
Default:
4.2.3.4.5.6 <correlations>
Description: These are the posterior correlations between vp, vs and density after inversion.
Argument: yes or no
Default:
4.2.3.4.5.7 <seismic-quality-grid>
Description: Quality grid kriged from values for t between facies probabilities and facies ob-
served in each of the wells.
Argument: yes or no
Default:
4.2.3.5 <well-output>
Description: Collects all output that can be given in well format. Wells contain logs for vp, vs
and density, each of these in four versions: Raw, ltered to background frequency, ltered to
seismic frequency and seismic resolution. In addition, the facies log is written if found.
Argument: Elements collecting output given in well format
Default:
CRAVA User Manual version 1.2 60
4.2.3.5.1 <format>
Description: Controls well formats used for output. Default is RMS.
Argument: Elements controlling the formats used for output
Default:
4.2.3.5.1.1 <rms>
Description: Controls if wells are written on RMS format.
Argument: yes or no
Default:
4.2.3.5.1.2 <norsar>
Description: Controls if wells are written on NORSAR format.
Argument: yes or no
Default:
4.2.3.5.2 <wells>
Description: Writes wells following original sampling density.
Argument: yes or no
Default:
4.2.3.5.3 <blocked-wells>
Description: Writes wells sampled to internal grid resolution.
Argument: yes or no
Default:
4.2.3.5.4 <blocked-logs>
Description: Not currently active.
Argument:
Default:
4.2.3.6 <wavelet-output>
Description: Collects all output that can be given for wavelets.
Argument: Elements controlling wavelet output
Default:
4.2.3.6.1 <format>
Description: Controls wavelet formats used for output. Default is JASON.
Argument: Elements controlling the formats used for output
Default:
4.2.3.6.1.1 <jason>
Description: Controls if wavelets are written on JASON form, being wlt format.
Argument: yes or no
Default: yes
CRAVA User Manual version 1.2 61
4.2.3.6.1.2 <norsar>
Description: Controls if wavelets are written on NORSAR form, being swav format.
Argument: yes or no
Default: no
4.2.3.6.2 <well-wavelets>
Description: Writes estimated wavelets for each well used for wavelet estimation. Note: These
wavelets are not optimally shifted, but aligned for easy comparison.
Argument: yes or no
Default: no
4.2.3.6.3 <global-wavelets>
Description: Writes global wavelets for each seismic angle.
Argument: yes or no
Default: no
4.2.3.6.4 <local-wavelets>
Description: Writes estimated local wavelet shift and scale surfaces. Can only be written when
<local-wavelet> is requested.
Argument: yes or no
Default:
4.2.3.7 <other-output>
Description: Controls output that is neither standard grid nor well.
Argument: Elements controlling output
Default:
4.2.3.7.1 <extra-surfaces>
Description: Temporary, will be replaced. Currently writes:
Top and base surface for constant thickness interval used for log ltering and facies
probabilities.
Top and base surface for extended inversion interval computed fromcorrelation surface.
Top and base surface for background estimation interval (larger than inversion interval).
Argument: yes or no
Default:
4.2.3.7.2 <prior-correlations>
Description: Write prior correlation les.
Argument: yes or no
Default:
4.2.3.7.3 <background-trend-1d>
Description: Write the background trend as 1D curve.
Argument: yes or no
CRAVA User Manual version 1.2 62
Default:
4.2.3.7.4 <local-noise>
Description: Writes estimated local noise surface. Can only be written when <local-noise-scaled>
or <estimate-local-noise> is requested.
Argument: yes or no
Default:
4.2.3.7.5 <rock-physics-distributions>
Description: Writes rock physics distribution per facies, in a vp, vs, and density STORM-grid. Only
available when facies probabilities are computed.
Argument: yes or no
Default: no
4.2.3.7.6 <error-file>
Description: Writes all errors to a separate le, in addition to the log le.
Argument: yes or no
Default: no
4.2.3.7.7 <task-file>
Description: Writes all tasks to a separate le, in addition to the log le.
Argument: yes or no
Default: no
4.2.3.8 <file-output-prefix>
Description: Common prex added to all les written in the run. Identies the run.
Argument: String
Default:
4.2.3.9 <log-level>
Description:
Argument: String. Possible values are error, warning, low, medium, high.
Default: Low
4.2.4 <advanced-settings>
Description: A collection of different commands that control advanced aspects of the program
control.
Argument: Commands controlling advanced aspects of the program control
Default:
4.2.4.1 <fft-grid-padding>
Description: Controls the padding size, can be used to optimize memory or improve visual results.
Padding should be at least one range laterally, and a wavelet length vertically to avoid edge
effects.
Argument: Elements controlling the padding size
Default:
CRAVA User Manual version 1.2 63
4.2.4.1.1 <x-fraction>
Description: Value telling how large the padding in the x-direction should be relative to the x-
length.
Argument: Value
Default: 0.0
4.2.4.1.2 <y-fraction>
Description: Value telling how large the padding in the x-direction should be relative to the y-
length.
Argument: Value
Default: 0.0
4.2.4.1.3 <z-fraction>
Description: Value telling how large the padding in the x-direction should be relative to the thick-
ness.
Argument: Value
Default: 0.0
4.2.4.2 <use-intermediate-disk-storage>
Description: When running under Windows with less physical memory than the program re-
quires, this activates a built-in smart-swap. It is more efcient to use this smart-swapping
than the built-in windows paging system. Linux/Unix swap is so efcient that this option
has little effect there. If you run Crava on a machine that you share with other users, it can be
wise to use this if you know that Crava will need most of the memory.
Argument: yes or no
Default:
4.2.4.3 <vp-vs-ratio>
Description: Value of Vp/Vs ratio used in reection matrix. By default, the Vp/Vs ratio is esti-
mated from the background model.
Argument: Value
Default: Not set
4.2.4.4 <vp-vs-ratio-from-wells>
Description: If this command is given, the Vp/Vs ratio used in the reection matrix will be esti-
mated from well data. By default, the ratio is taken from the background model. If the key-
word <wavelet-estimation-interval> has also been specied, the estimate will be limited
to that interval.
Argument: yes or no
Default: no
4.2.4.5 <maximum-relative-thickness-difference>
Description: Value giving the limit of how small the minimum interval thickness can be relative to
maximum. If this gets too low, the transformation to stationarity for the FFT-algorithm gives
strange results.
Argument: Value
Default: Default is 0.5, which is acceptable. Slightly smaller seems to work as well.
CRAVA User Manual version 1.2 64
4.2.4.6 <frequency-band>
Description: This command controls the frequency band of the inversion, so high and/or low
frequencies can be ltered away. This ought to be done by the wavelet, but can be done here.
Argument: Elements controlling the frequency band of the inversion
Default:
4.2.4.6.1 <low-cut>
Description: Value setting the minimum frequency affected by the inversion.
Argument: Value
Default: 5.0
4.2.4.6.2 <high-cut>
Description: Value setting the maximum frequency affected by the inversion.
Argument: Value
Default: 55.0
4.2.4.7 <energy-threshold>
Description: If the energy in a trace falls below this threshold relative to the average, the trace is
interpolated from neighbours.
Argument: Value
Default: 0.0
4.2.4.8 <wavelet-tapering-length>
Description: Value giving the length of the wavelet to be estimated in ms.
Argument: Value
Default: 200.0
4.2.4.9 <minimum-relative-wavelet-amplitude>
Description: Value giving the ratio between the smallest relevant amplitude and the largest am-
plitude of peaks on an estimated wavelet. Edge peaks below this ratio are removed.
Argument: Value
Default: 0.05
4.2.4.10 <maximum-wavelet-shift>
Description: Value controlling how much the wavelet is allowed to be shifted when doing estima-
tion of wavelet or noise.
Argument: Value
Default: 11.0
4.2.4.11 <minimum-sampling-density>
Description: Threshold value for minimum sampling density allowed.
Argument: Value
Default: 0.5 ms
4.2.4.12 <minimum-horizontal-resolution>
Description: Threshold value for minimum horizontal resolution allowed.
Argument: Value
Default: 5 m
CRAVA User Manual version 1.2 65
4.2.4.13 <white-noise-component>
Description: In order to stabilise the inversion, we need to interpret some of the noise as white.
This value controls the fraction.
Argument: Value between 0 and 1.
Default: 0.1
4.2.4.14 <reflection-matrix>
Description: The le should be a 3 by number of seismic data cubes ascii matrix. The rst column is
the factor used for vp for each cube setting (angle and ps/pp) when computing the reection
coefcients. The second and third are for vs and density.
Argument: File name
Default: Linearised Aki-Richards.
4.2.4.15 <kriging-data-limit>
Description: Integer value giving the limit for the amount of well data used to krige each point. A
high value gives a smoother and more exact eld, but takes more time.
Argument: Integer
Default: 250
4.2.4.16 <debug-level>
Description: Gives debug messages and output. Not intended for use except on request by NR.
Argument: Integer value 0, 1 or 2.
Default: 0
4.2.4.17 <smooth-kriged-parameters>
Description: Tells whether we should smooth borders between kriging blocks or not.
Argument: yes or no
Default: no
4.2.4.18 <rms-panel-mode>
Description: Disables some checks that are unnecessary when running from RMS.
Argument: yes or no
Default: no
4.2.4.19 <guard-zone>
Description: Changes the amount of data that will be required outside the interval of interest.
The guard zone contains data that will contribute to the inversion results in the interval of
interest, and reducing the guard zone is therefore discouraged.
Argument: Value
Default: 100ms
4.2.4.20 <3d-wavelet-tuning-factor>
Description: Allows tuning of the 3D wavelet estimation. A large value forces better t of wavelet,
while a smaller value gives smaller secondary peaks in wavelet.
Argument: Value
Default: 50
CRAVA User Manual version 1.2 66
4.2.4.21 <gradient-smoothing-range>
Description: Controls smoothing of the gradient used in 3D wavelet estimate and 3D inversion.
This should be of the order of the horisontal extent of the 3D wavelet.
Argument: Value
Default: 100
4.2.4.22 <estimate-well-gradient-from-seismic>
Description: If yes, the well gradient used for 3D wavelet estimation is estimated from seismic.
Otherwise this is taken from the correlation direction.
Argument: yes or no
Default: no
4.3 <survey> (necessary)
Description: All information about the seismic data is collected here.
Argument: Elements containing information about the seismic data
Default:
4.3.1 <angular-correlation>
Description: 1D variogram. Gives the noise correlation between survey angles.
Argument: 1D variogram, see Section 4.6.
Default:
4.3.2 <segy-start-time>
Description: Global start time for segy cubes. This is used if no individual time is given for a segy-
cube.
Argument: Value
Default:
4.3.3 <angle-gather> (necessary)
Description: Repeatable command, one for each seismic data cube.
Argument: Elements containing information about the different seismic data cubes
Default:
4.3.3.1 <offset-angle> (necessary)
Description: This is the angle for the seismic data cube.
Argument: Value
Default:
4.3.3.2 <seismic-data> (necessary)
Description: Information about the seismic data cube.
Argument: Elements containing information about the seismic data cube
Default:
4.3.3.2.1 <file-name> (necessary)
Description: File name for the seismic data cube of one of the standard 3D formats, see Sec-
tion 3.1.8. The le type will be automatically detected.
Argument: File name
Default:
CRAVA User Manual version 1.2 67
4.3.3.2.2 <start-time>
Description: Value giving the start time for this segy le.
Argument: Value
Default:
4.3.3.2.3 <segy-format>
Description: Information about the segy format. By default, CRAVA recognises SeisWorks, IESX,
SIP and Charisma. See Table 3.1.
Argument: Elements containing information about the segy format
Default:
4.3.3.2.3.1 <standard-format>
Description: Giving the starting format for modications.
Argument: seisworks, iesx, charisma or SIP
Default: seisworks
4.3.3.2.3.2 <location-x>
Description: The byte location for the x-coordinate in the trace header.
Argument: Integer
Default:
4.3.3.2.3.3 <location-y>
Description: The byte location for the y-coordinate in the trace header.
Argument: Integer
Default:
4.3.3.2.3.4 <location-il>
Description: The byte location for the inline in the trace header.
Argument: Integer
Default:
4.3.3.2.3.5 <location-xl>
Description: The byte location for the crossline in the trace header.
Argument: Integer
Default:
4.3.3.2.3.6 <bypass-coordinate-scaling>
Description: Indicates whether coordinate scaling information should be used.
Argument: yes or no
Default:
4.3.3.2.3.7 <location-scaling-coefficient>
Description:
Argument: Integer
CRAVA User Manual version 1.2 68
Default:
4.3.3.2.4 <type>
Description: Indicating the type of seismic data. Note that if both pp and ps cubes are used, these
must be aligned.
Argument: pp or ps
Default: pp
4.3.3.3 <wavelet>
Description: Information about the wavelet for this angle and seismic type. If not given, the wavelet
will be estimated.
Argument: Elements containing information about the wavelet and seismic type
Default:
4.3.3.3.1 <file-name>
Description: File name for wavelet le on JASON or NORSAR format. Can not be given together
with ricker. If neither le-name nor ricker is given, wavelet is estimated.
Argument: File name
Default:
4.3.3.3.2 <ricker>
Description: Use Ricker wavelet. Can not be given together with le-name. If neither le-name
nor ricker is given, wavelet is estimated.
Argument: Peak frequency
Default:
4.3.3.3.3 <scale>
Description: Wavelet read from le or ricker wavelet is multiplied by this. Has no meaning when
wavelet is estimated.
Argument: Value
Default:
4.3.3.3.4 <estimate-scale>
Description: Should global scale be estimated?
Argument: yes or no
Default:
4.3.3.3.5 <local-wavelet>
Description: The amplitude and shift of the wavelet may be modied locally by 2D elds for shift
and scale values. This is handled here.
Argument: Elements modifying the amplitude and shift of the wavelet
Default:
4.3.3.3.5.1 <shift-file>
CRAVA User Manual version 1.2 69
Description: File name for standard surface le giving the local shift for the wavelet. Not allowed
when wavelet is estimated.
Argument: File name
Default:
4.3.3.3.5.2 <scale-file>
Description: File name for standard surface le giving local scale for the wavelet. Not allowed
when wavelet is estimated.
Argument: File name
Default:
4.3.3.3.5.3 <estimate-shift>
Description: Should a local shift be estimated? Not allowed with <shift-file>, but can be used
both with given and estimated wavelet.
Argument: yes or no
Default:
4.3.3.3.5.4 <estimate-scale>
Description: Should a local scale be estimated? Not allowed with <scale-file>, but can be used
both with given and estimated wavelet.
Argument: yes or no
Default:
4.3.3.4 <wavelet-3d>
Description: Information about the 3D-wavelet for this angle and seismic type. If not given, a 1D-
wavelet is assumed.
Argument: Elements containing information about the 3D-wavelet
Default:
4.3.3.4.1 <file-name>
Description: File name for the 1D-wavelet le. This 1D-wavelet will dene the 3D-wavelet in
combination with the lter given in <processing-factor-file-name>. If not given, the 1D-
wavelet is estimated.
Argument: File name
Default:
4.3.3.4.2 <processing-factor-file-name>
Description: File name for 3D-wavelet damping factor lter le. The 3D-wavelet is dened by the
1D-wavelet and the lter. The 1D-wavelet is either given in <file-name> or estimated. In
either case this lter le must be given.
Argument: File name for the amplitude scalings in the wavenumber lter.
Default:
4.3.3.4.3 <propagation-factor-file-name>
Description: File name for 3D-wavelet correction lter le. This is used to set up the noise model
for the 3D-wavelet.
CRAVA User Manual version 1.2 70
Argument: File name for the correction factors in the wavenumber lter.
Default:
4.3.3.4.4 <stretch-factor>
Description: Stretch factor for 3D-wavelet. The pulse is stretch with this factor.
Argument: Value > 0.0
Default: 1.0
4.3.3.4.5 <estimation-range-x-direction>
Description: Range for area around the well in x-direction where data are used in 3D wavelet
estimation.
Argument: Value >= 0.0
Default: 0.0
4.3.3.4.6 <estimation-range-y-direction>
Description: Range for area around the well in y-direction where data are used in 3D wavelet
estimation.
Argument: Value >= 0.0
Default: 0.0
4.3.3.5 <match-energies>
Description: If yes, signal to noise ratio and wavelet scaling will be set to match model values
with empirical values. Not a common estimator.
Argument: yes or no
Default:
4.3.3.6 <signal-to-noise-ratio>
Description: Value for the signal to noise value. If not given, this will be estimated.
Argument: Value
Default:
4.3.3.7 <local-noise-scaled>
Description: Name of standard surface le with local noise.
Argument: File name
Default:
4.3.3.8 <estimate-local-noise>
Description: Can not say yes here if <local-noise-scaled> is given.
Argument: yes or no
Default:
4.3.4 <wavelet-estimation-interval>
Description: Controls the time interval used for wavelet estimation by a top and base surface.
Argument: Elements controlling the time interval used for wavelet estimation
Default: By default, estimation is done from all available seismic and well data.
CRAVA User Manual version 1.2 71
4.3.4.1 <top-surface-file>
Description: File name for standard surface le giving the top of the time interval used for wavelet
estimation.
Argument: File name
Default:
4.3.4.2 <base-surface-file>
Description: File name for standard surface le giving the base of the time interval used for wavelet
estimation.
Argument: File name
Default:
4.3.5 <time-gradient-settings>
Description: The prior standard deviation of the gradients are given, as well as the minimum
distance for where gradient lines should not cross each other
Argument: The standard deviation and the minimum distance
Default:
4.3.5.1 <distance>
Description: The minimum lateral distance for where the gradient lines should not cross. The dis-
tance is equal for both x- and y-direction.
Argument: Value
Default: 100 m
4.3.5.2 <sigma>
Description: The prior standard deviation of the gradient. Equal for both x- and y-gradients
Argument: Value
Default: 1 ms/m
4.4 <well-data>
Description: All information about the well data is collected here.
Argument: Elements containing information about the well data
Default:
4.4.1 <log-names>
Description: CRAVA needs to nd the time, vp, vs, density and possibly facies logs. The name of
these logs in the well les are given here.
Argument: Name of logs
Default:
4.4.1.1 <time>
Description: Name of the TWT log
Argument: String
Default:
4.4.1.2 <vp>
Description: Name of the vp log. May not be given if <dt> is given.
Argument: String
CRAVA User Manual version 1.2 72
Default:
4.4.1.3 <dt>
Description: Name of the inverse vp log. May not be given if <vp> is given.
Argument: String
Default:
4.4.1.4 <vs>
Description: Name of the vs log. May not be given if <dts> is given.
Argument: String
Default:
4.4.1.5 <dts>
Description: Name of the inverse vs log. May not be given if <vs> is given.
Argument: String
Default:
4.4.1.6 <density>
Description: Name of the density log.
Argument: String
Default:
4.4.1.7 <facies>
Description: Name of the facies log.
Argument: String
Default:
4.4.2 <well>
Description: Repeatable command, one for each well. Contains information about the wells.
Argument: Elements containing information about the wells
Default:
4.4.2.1 <file-name>
Description: File name for a well le. RMS or Norsar format.
Argument: File name
Default:
4.4.2.2 <use-for-wavelet-estimation>
Description: Should this well be used for wavelet estimation?
Argument: yes or no
Default:
4.4.2.3 <use-for-background-trend>
Description: Should this well be used for background trend estimation?
Argument: yes or no
Default:
4.4.2.4 <use-for-facies-probabilities>
Description: Should this well be used for facies probability estimation?
CRAVA User Manual version 1.2 73
Argument: yes or no
Default:
4.4.2.5 <synthetic-vs-log>
Description: Is the Vs log in this well synthetic? Will be detected from Vp correlation if not speci-
ed here.
Argument: yes or no
Default:
4.4.2.6 <filter-elastic-logs>
Description: Should we multi-parameter-lter the elastic logs in this well after the inversion?
Argument: yes or no
Default:
4.4.2.7 <optimize-position>
Description: Repeatable command, one for each offset angle used for estimating optimised well
location for this well.
Argument: Elements controlling optimisation of well location
Default:
4.4.2.7.1 <angle>
Description: Offset angle used for estimating optimised well location
Argument: Value
Default:
4.4.2.7.2 <weight>
Description: Weight of the offset angle given in <angle>
Argument: Value
Default: 1
4.4.3 <high-cut-seismic-resolution>
Description: This frequency is used to lter wells down to seismic resolution. Only used to gener-
ate output logs for QC.
Argument: Value
Default:
4.4.4 <allowed-parameter-values>
Description: Sometimes there are faulty values in well logs. Here, trigger values for error detection
can be controlled. These fall in two categories: Actual log values that are wrong, or logs that
have extremely low or high variance when the background model is subtracted.
Argument: Elements controlling trigger values for error detection
Default:
4.4.4.1 <minimum-vp>
Description: Value for the smallest legal vp value.
Argument: Value
Default: 1300 m/s
CRAVA User Manual version 1.2 74
4.4.4.2 <maximum-vp>
Description: Value for the largest legal vp value.
Argument: Value
Default: 7000 m/s
4.4.4.3 <minimum-vs>
Description: Value for the smallest legal vs value.
Argument: Value
Default: 200 m/s
4.4.4.4 <maximum-vs>
Description: Value for the largest legal vs value.
Argument: Value
Default: 4200 m/s
4.4.4.5 <minimum-density>
Description: Value for the smallest legal density value.
Argument: Value
Default: 1.4 g/cm
3
4.4.4.6 <maximum-density>
Description: Value for the largest legal density value.
Argument: Value
Default: 3.3 g/cm
3
4.4.4.7 <minimum-variance-vp>
Description: Value for the smallest legal variance in the vp log after the background is subtracted
and logarithm is taken.
Argument: Value
Default: 0.0005
4.4.4.8 <maximum-variance-vp>
Description: Value for the largest legal variance in the vp log after the background is subtracted
and logarithm is taken.
Argument: Value
Default: 0.0250
4.4.4.9 <minimum-variance-vs>
Description: Value for the smallest legal variance in the vs log after the background is subtracted
and logarithm is taken.
Argument: Value
Default: 0.0010
4.4.4.10 <maximum-variance-vs>
Description: Value for the largest legal variance in the vs log after the background is subtracted
and logarithm is taken.
Argument: Value
Default: 0.0500
CRAVA User Manual version 1.2 75
4.4.4.11 <minimum-variance-density>
Description: Value for the smallest legal variance in the density log after the background is sub-
tracted and logarithm is taken.
Argument: Value
Default: 0.0002
4.4.4.12 <maximum-variance-density>
Description: Value for the largest legal variance in the density log after the background is sub-
tracted and logarithm is taken.
Argument: Value
Default: 0.0100
4.4.4.13 <minimum-vp-vs-ratio>
Description: Value for the smallest Vp/Vs-ratio regarded as likely.
Argument: Value
Default: 1.4
4.4.4.14 <maximum-vp-vs-ratio>
Description: Value for the largest Vp/Vs-ratio regarded as likely.
Argument: Value
Default: 3.0
4.4.5 <maximum-deviation-angle>
Description: Value for the maximumdeviation angle of a well before it is excluded fromestimation
based on vertical wells (such as wavelet and signal to noise).
Argument: Value
Default: 15
4.4.6 <maximum-rank-correlation>
Description: If the correlation between vp and vs logs exceed this value, the vs log is considered
to be synthetic, and not counted as additional data in estimation.
Argument: Value close to, but less than 1.
Default: 0.99
4.4.7 <maximum-merge-distance>
Description: Value giving the minimum distance in time between well log entries before they are
merged to one observation.
Argument: Value
Default: 0.01
4.4.8 <maximum-offset>
Description: Value giving the maximum allowed offset for moving wells in meters.
Argument: Value
Default: 250
4.4.9 <maximum-shift>
Description: Value giving the maximum allowed vertical shift for moving wells.
Argument: Value
Default: 11.0
CRAVA User Manual version 1.2 76
4.4.10 <well-move-data-interval>
Description: Denes an interval for estimation of facies probability given elastic parameters.
Argument: Elements dening estimation interval
Default: Everywhere facies and elastic logs are present.
4.4.10.1 <top-surface-file>
Description: File name for standard surface le giving the top of the estimation interval.
Argument: File name
Default:
4.4.10.2 <base-surface-file>
Description: File name for standard surface le giving the base of the estimation interval.
Argument: File name
Default:
4.5 <prior-model>
Description: This command denes the prior model for elastic parameters and possibly also facies.
Argument: Elements dening prior models for elastic parameters and facies
Default:
4.5.1 <background>
Description: Contains information about the background model or how to estimate it. Note that
either all parameters must be given, or all must be estimated.
Argument: Elements containing information about the background model
Default:
4.5.1.1 <ai-file>
Description: File name for 3D grid le, giving background AI. Can not be given together with
<vp-file> or <vp-constant>.
Argument: File name
Default:
4.5.1.2 <si-file>
Description: File name for 3D grid le, giving background SI. Can not be given together with
<vs-file> or <vs-constant> or <vp-vs-ratio-file>.
Argument: File name
Default:
4.5.1.3 <vp-vs-ratio-file>
Description: File name for 3D grid le, giving background Vp/Vs. Can not be given together with
<vs-file> or <vs-constant>.
Argument: File name
Default:
4.5.1.4 <vp-file>
Description: File name for 3D grid le, giving background vp. Can not be given together with
<vp-constant>.
Argument: File name
CRAVA User Manual version 1.2 77
Default:
4.5.1.5 <vs-file>
Description: File name for 3D grid le, giving background vs. Can not be given together with
<vs-constant>.
Argument: File name
Default:
4.5.1.6 <density-file>
Description: File name for 3D grid le, giving background density. Can not be given together with
<density-constant>.
Argument: File name
Default:
4.5.1.7 <vp-constant>
Description: Value, used for constant vp background. Can not be given together with <vp-file>.
Argument: Value
Default:
4.5.1.8 <vs-constant>
Description: Value, used for constant vs background. Can not be given together with <vs-file>.
Argument: Value
Default:
4.5.1.9 <density-constant>
Description: Value, used for constant density background. Can not be given together with <density-file>.
Argument: Value
Default:
4.5.1.10 <velocity-field>
Description: File name for 3D grid le giving a velocity eld used as base for vp in background
estimation. Can not be used if the background parameters are given.
Argument: File name
Default:
4.5.1.11 <lateral-correlation>
Description: 2D variogram for the lateral correlation in the elastic parameters in the estimated
background model, used for kriging of wells. Can not be used if the background parameters
are given.
Argument: 2D variogram, see Section 4.6.
Default:
4.5.1.12 <high-cut-background-modelling>
Description: Value giving the maximum frequency in the estimated background model. Can not
be used if the background parameters are given.
Argument: Value
Default: 6.0 Hz
CRAVA User Manual version 1.2 78
4.5.2 <earth-model>
Description: Contains inverted seismic data used for forward modelling.
Argument: Vp, vs and rho used for forward modelling.
Default:
4.5.2.1 <vp-file>
Description: File name for 3D grid le, giving vp.
Argument: File name
Default:
4.5.2.2 <vs-file>
Description: File name for 3D grid le, giving vs.
Argument: File name
Default:
4.5.2.3 <density-file>
Description: File name 3D grid le, giving density.
Argument: File name
Default:
4.5.2.4 <ai-file>
Description: File name for 3D grid le, giving AI. Can not be given together with <vp-file> .
Argument: File name
Default:
4.5.2.5 <si-file>
Description: File name for 3D grid le, giving SI. Can not be given together with <vs-file2> or
<vp-vs-ratio-file2>.
Argument: File name
Default:
4.5.2.6 <vp-vs-ratio-file>
Description: File name for 3Dgrid le, giving Vp/Vs. Can not be given together with <vs-file2>.
Argument: File name
Default:
4.5.3 <local-wavelet>
Description: Contains prior information for local wavelet modelling.
Argument: Elements containing prior information for local wavelet estimation.
Default:
4.5.3.1 <lateral-correlation>
Description: 2D variogram for the lateral correlation in local wavelet modelling.
Argument: 2D variogram, see Section 4.6.
Default:
4.5.4 <lateral-correlation>
Description: 2D variogram for the lateral correlation in the elastic parameters.
Argument: 2D variogram, see Section 4.6.
CRAVA User Manual version 1.2 79
Default:
4.5.5 <temporal-correlation>
Description: File name for the temporal correlation le. The le is an ascii le. Usually, this le
comes from an earlier run of CRAVA.
Argument: File name
Default:
4.5.6 <parameter-correlation>
Description: File name for the parameter correlation le. The le is an ascii le, containing covari-
ances between Vp, Vs and density. It is most common to use a le resulting from an earlier
run of CRAVA, and the le might look like this:
0.001171 0.000786 0.000046
0.000786 0.001810 -0.000395
0.000046 -0.000395 0.000530
Argument: File name
Default:
4.5.7 <correlation-direction>
Description: File name for the standard surface le giving the correlation direction for the inver-
sion.
Argument: File name
Default:
4.5.8 <facies-probabilities>
Description: Commands controlling the generating of facies probabilities.
Argument: Elements controlling facies probabilities
Default:
4.5.8.1 <use-vs>
Description: Decides whether V
s
information is used when computing facies probabilities.
Argument: yes or no.
Default: yes.
4.5.8.2 <use-prediction>
Description: Decides whether sampled inversion logs are used when computing facies probabili-
ties. If not, ltered logs are used.
Argument: yes or no.
Default: no.
4.5.8.3 <use-absolute-elastic-parameters>
Description: Decides whether facies probabilities are generated based on absolute elastic parame-
ters or elastic parameters minus trend (background model).
Argument: yes or no
Default: no
4.5.8.4 <estimation-interval>
Description: Denes an interval for estimation of facies probability given elastic parameters.
CRAVA User Manual version 1.2 80
Argument: Elements dening estimation interval
Default: Everywhere facies and elastic logs are present.
4.5.8.4.1 <top-surface-file>
Description: File name for standard surface le giving the top of the estimation interval.
Argument: File name
Default:
4.5.8.4.2 <base-surface-file>
Description: File name for standard surface le giving the base of the estimation interval.
Argument: File name
Default:
4.5.8.5 <prior-probabilities>
Description: Prior facies probabilities are given for all facies. Priors can be given as constant num-
bers or 3D cubes. If this command is not given, prior distribution is estimated from wells.
Argument: Elements controlling facies probabilities
Default:
4.5.8.5.1 <facies>
Description: Repeatable command, one for each facies. All facies present in well logs must be
given.
Argument: Elements containing information about the facies
Default:
4.5.8.5.1.1 <name>
Description: Name of facies.
Argument: Facies name
Default:
4.5.8.5.1.2 <probability>
Description: Probability for the facies given above. Either this command or <probability-cube>
is given, same for all facies.
Argument: Real numbers between 0 and 1. Numbers for all facies must sum to one.
Default:
4.5.8.5.1.3 <probability-cube>
Description: File name for 3D grid le containing prior facies probability for facies with name
given above. Either this command or <probability> is given, same for all facies.
Argument: File name
Default:
4.5.8.6 <uncertainty-level>
Description: Value dening how large the undened probability will be when facies probabilities
are computed. This value is scaled and used as likelihood for undened when facies proba-
bilities are computed.
CRAVA User Manual version 1.2 81
Argument: Value
Default: 0.01
4.6 Variogram
The variograms are given on the following form:
4.6.1 <variogram-type>
Description: Either genexp or spherical for general exponential or spherical variogram.
Argument:
Default:
4.6.2 <angle>
Description: Value for the azimuth direction. Only for 2D variograms.
Argument:
Default:
4.6.3 <range>
Description: Value for the range in the azimuth direction.
Argument:
Default:
4.6.4 <subrange>
Description: Value for the range normal to the azimuth direction. Only for 2D variograms.
Argument:
Default:
4.6.5 <power>
Description: Value between 1 and 2 for the power of the general exponential variogram. Not al-
lowed for spherical variogram.
Argument:
Default:
All angles are given as mathematical angles in degrees.
CRAVA User Manual version 1.2 82
A Sample model le
<crava>
<actions>
<mode> inversion </mode>
<inversion-settings>
<prediction> yes </prediction>
</inversion-settings>
</actions>
<survey>
<angular-correlation>
<variogram-type> genexp </variogram-type>
<range> 10 </range>
<power> 1 </power>
</angular-correlation>
<segy-start-time> 2500 </segy-start-time>
<angle-gather>
<offset-angle> 16 </offset-angle>
<seismic-data>
<file-name> seismic/Cube16.segy </file-name>
</seismic-data>
</angle-gather>
<angle-gather>
<offset-angle> 28 </offset-angle>
<seismic-data>
<file-name> seismic/Cube28.segy </file-name>
</seismic-data>
</angle-gather>
</survey>
<well-data>
<log-names>
<time> TWT </time>
<dt> DT </dt>
<dts> DTS </dts>
<density> RHOB </density>
</log-names>
<well>
<file-name> logs/ed6406_3-3_cut.rms </file-name>
</well>
<well>
<file-name> logs/ed6506_12-3_cut.rms </file-name>
</well>
<well>
<file-name> logs/ed6506_12-8_cut.rms </file-name>
</well>
<well>
<file-name> logs/ed6506_12-S4H_cut.rms </file-name>
</well>
</well-data>
<prior-model>
CRAVA User Manual version 1.2 83
<background>
<high-cut-background-modelling> 6 </high-cut-background-modelling>
<lateral-correlation>
<variogram-type> genexp </variogram-type>
<range> 2000 </range>
<subrange> 1000 </subrange>
<angle> 70 </angle>
<power> 1 </power>
</lateral-correlation>
</background>
<lateral-correlation>
<variogram-type> genexp </variogram-type>
<range> 1000 </range>
<subrange> 500 </subrange>
<angle> 70 </angle>
<power> 1 </power>
</lateral-correlation>
</prior-model>
<project-settings>
<output-volume>
<utm-coordinates>
<reference-point-x> 398500 </reference-point-x>
<reference-point-y> 7210300 </reference-point-y>
<length-x> 4500 </length-x>
<length-y> 4500 </length-y>
<angle> 23.627 </angle>
<sample-density-x> 50 </sample-density-x>
<sample-density-y> 50 </sample-density-y>
</utm-coordinates>
<interval-two-surfaces>
<top-surface>
<time-file> horizons/TopTime_avg3100ms.storm </time-file>
<depth-file> horizons/TopTime_avg3100ms.storm </depth-file>
</top-surface>
<base-surface>
<time-file> horizons/BaseTime_avg3600ms.storm </time-file>
</base-surface>
<velocity-field-from-inversion> yes </velocity-field-from-inversion>
<number-of-layers> 140 </number-of-layers>
</interval-two-surfaces>
</output-volume>
<io-settings>
<log-level> high </log-level>
<top-directory> ../ </top-directory>
<input-directory> input/ </input-directory>
<output-directory> output/ </output-directory>
<grid-output>
<format>
<ascii> yes </ascii>
</format>
CRAVA User Manual version 1.2 84
<domain>
<time> yes </time>
<depth> yes </depth>
</domain>
<elastic-parameters>
<vp> yes </vp>
<vs> yes </vs>
<density> yes </density>
<background> yes </background>
</elastic-parameters>
</grid-output>
<well-output>
<wells> yes </wells>
<blocked-wells> yes </blocked-wells>
</well-output>
<wavelet-output>
<format>
<jason> yes </jason>
</format>
<well-wavelets> yes </well-wavelets>
<global-wavelets> yes </global-wavelets>
</wavelet-output>
<other-output>
<prior-correlations> yes </prior-correlations>
<background-trend-1d> yes </background-trend-1d>
</other-output>
</io-settings>
<advanced-settings>
<fft-grid-padding>
<x-fraction> 0.200001 </x-fraction>
<y-fraction> 0.200001 </y-fraction>
<z-fraction> 1.000000 </z-fraction>
</fft-grid-padding>
</advanced-settings>
</project-settings>
</crava>
CRAVA User Manual version 1.2 85
B Test suite overview
This appendix gives an overview of the features that are currently tested in the test suite.
Test \ Case number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Mode
Forward x
Estimation - 1D x x
Estimation - 3D x
Inversion - prediction x x x x x x x x x x x x x
Inversion - simulation x
Inversion - kriging x x x
Inversion - facies prob. x x x x
Survey
Angle gathers 2 2 2 2 2 2 2 2 2 2 3 3 3 3 3 2 2 2
Estimate global wavelet x x x x x
Estimate global SN-ratio x x x x x
Estimate local noise x
Estimate local scale x
Estimate local shift x
Wavelet estimation interval x
Wells
Number of wells 1 1 1 4 4 4 4 1 1 6 6 6 6 1 1 2 3
Only synthetic Vs logs x x
<use-for-facies-probabilities> x x
<synthetic-vs-log> x
<lter-elastic-logs> x x
<optimize-position> x
<allowed-parameter-values> x x x x
Prior model
Estimate covariance x x x x x x x x
Estimate background x x x x x x x
Anisotropic background mod x x x x x
Anisotropic lateral corr x x x x x x x x
Correlation direction x
VpVsRho lter x x
VpRho lter x
Use prediction (not lter) x
Absolute params. for facies x
Facies estimation interval x x x
Output volume
Area from UTM x x x x x x x x x
Area from ILXL x
Area from surface
Area from rst angle gather x x x x x x x
One surface x
Two surfaces x x x x x x x x x x x x x x x
Constant top and base x
CRAVA User Manual version 1.2 86
Test \ Case number 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
Depth conversion
Vel. eld - from le x x
Vel. eld - from inversion x
Vel. eld - from surfaces
Number of depth surfaces 2 1 2
Advanced settings
FFT-grid-padding estimated x
Intermediate disk storage x
Smooth kriged parameters x
Well input formats
Roxar x x x x x x x x x x x x x x x x
Norsar x x
Surface input formats
STORM BINARY x x x x x x x x x x x x x
Roxar ASCII (Irap classic) x x x x x x x
Grid input formats
STORM BINARY x x x x
SEGY x x x x x x x x x
CRAVA x x x x x x
SGRI x x
Output check of
Depth domain x x
Generated seismic data x
Synthetic seismic data x x x
Pred/sim elastic parameters x x x x x x x x x
Background x x x x
Background trend x x x x
Prior correlations x x x
Posterior correlations x
Blocked wells x x x x x x x x x x x
Surfaces - top and base x x
Surfaces - help grids x
Wavelet - global x x x x x x
Wavelet - well x x x x x
Wavelet - local shape x
Wavelet - local scale/shift x
Local noise x
CRAVA User Manual version 1.2 87
C Release notes
The tags, e.g. CRA-49, are links to the CRAVA project management system called JIRA. Note that access
permission is required to be able to open the links.
C.1 Changes from v1.1 to v1.2
New features:
Introduce new and more efcient resampling of seismic data. CRA-274
Add SI (=Vs*Rho) as possible input type for background model. CRA-287
For forward modelling allowearth model to be input as (AI,SI,Rho) and (AI,Vp/Vs,Rho). CRA-302
Add new SegY format "SIPX" to list of auto-detected formats.CRA-290
Added 3D wavelet estimation/inversion. CRA-304, CRA-308, CRA-309
Changes:
Require seismic data in target zone+guard zone rather than in full inversion grid (simbox). CRA-
216
Cleaned test suite and ndgrammar for <TAB> and trailing blanks, and added "no trailings blanks"
as requirement for passing test suite. CRA-275, CRA-276
Use one wavelet (200ms) as default when estimating vertical padding size. Previous default was
half a wavelet (100ms). CRA-300
Ensure that we log the parameter values of all parameters available under the <advanced-settings>
keyword. CRA-307
Bug xes:
Fixed problem identifying the lateral geometry of seismic data. CRA-277
Vp and Vs were reported as input data types for background model in logFile.txt when AI and
Vp/Vs were used. CRA-288
An incorrect grid size estimate made inversion runs stop even for medium sized grids. CRA-292
Fixed crash when the <filter-elastic-logs> option was used. CRA-310
Removed memory leaks in test suite using Purify. CRA-265.
C.2 Changes from v1.0 to v1.1
New features:
Allow background model to be specied with AI and/or V
p
/V
s
, as a complement to V
p
, V
s
, and .
Specify les for these parameters with keyword <ai-file> and <vp-vs-ratio-file>. CRA-214
Allow users to specify V
p
/V
s
for use in the reection matrix. Use keyword <vp-vs-ratio>. CRA-
219
Remove certain QC checks when CRAVA is run from Irap RMS, as Irap RMS will perform these
tests. Activated with keyword <rms-panel-mode>. CRA-221
When seismic data is given in SegY format, write the IL-XL range of data to the log le. CRA-230,
CRA-267
Create synthetic residuals which add up to original seismic when added to synthetic seismic. Ac-
tivated with <synthetic-residuals>. CRA-234
CRAVA User Manual version 1.2 88
If the <rms-panel-mode> mode has been requested and facies probabilities are not requested, well
logs will not be (multi-parameter) ltered. CRA-242
Allow V
p
/V
s
to be estimated from wells. Activated with keyword <vp-vs-ratio-from-wells>. If
<wavelet-estimation-interval> has been specied, this interval will also be used for the V
p
/V
s
estimate. CRA-247
If automatic SegY format detection fails, list trace header locations of known formats as well as
their associated CRAVA-names. CRA-259
Give a warning if prior or posterior parameter correlations exceed one. To avoid this situation
completely, some recoding is required. This is left for a later release (see CRA-257). CRA-261
When specifying an area as UTM coordinates or surface, choose as inversion area the smallest
IL-XL box that encloses the specied area. Activated with keyword <snap-to-seismic-data>.
CRA-266
Allow (some) traces to have bogus IL-XL information in header. If encountered, the IL-XL values
are set to undened. CRA-269
Write IL-XL values for well start positions whenever possible. This makes it easy to do inversion
around a given well. CRA-270
Changes:
When the prior correlation between V
p
and V
s
cannot be estimated from well data it is set to
1/

2 0.7. CRA-220
Made blocking of facies deterministic. If a cell has the same facies count for n different facies, the
cell value becomes facies_to_choose = (k + 1) % n, where k is the layer number. This used to be
randomly chosen, so that identical runs could give different facies predictions. CRA-249
Changed the computation of SegY geometry, and improved error messages. The new version
should be more accurate and robust.
Bug xes:
Fixed crash in background modelling when there were (unrealistically) few layers in grid. CRA-
200
Background volumes of AI and/or V
p
/V
s
containing undened values caused troubles. The x is
to set such undened values to global mean. CRA-222
Removed bugs in estimation of wavelet norm, which indirectly led to ringing in inversion results.
CRA-223
Number of allocated grids were incorrectly counted when estimating total memory requirement
for CRAVA. CRA-251
Fixed crash when interval was specied with option <interval-one-surface>. Wavelets had not
been transformed prior to constant thickness inversion. Also, a wavelet polarity bug was xed.
New test cases 16 and 17 were added to avoid future constant thickness and wavelet polarity
related bugs. CRA-252, CRA-253
For short wells, deviation angles were incorrectly estimated due to an 10ms sample requirement.
Also, the angle was calculated between well start and well end instead of for trajectory tangents.
Fixed crash when project directory could not be created. This was a bug in the throw/catch system
of NRLib. CRA-256
Fixed several bugs encountered when using 3D prior facies probability volumes. CRA-272
Fixed bug with interpolation of Storm-grids used for input.
CRAVA User Manual version 1.2 89
References
Aki, K. and Richards, P. G. (1980). Quantitative Seismology: Theory and Methods. W. H. Freeman and Co. 15
Brown, R. L. (1992). Estimation of AVO in the prescence of noise using prestack migration. pages 855888.
14
Buland, A., Kolbjrnsen, O., and Omre, H. (2003). Rapid spatially coupled AVO inversion in the fourier
domain. Geophysics, 68:824836. 14, 18, 28
Buland, A. and Landr, M. (2001). The impact of common offset migration on porosity estimation by AVO
inversion. Geophysics, 66:755762. 14
Buland, A., Landr, M., Andersen, M., and Dahl, T. (1996). AVO inversion of troll eld data. Geophysics,
61:15891602. 14
Buland, A. and Omre, H. (2003). Bayesian linearized AVO inversion. Geophysics, 68:185198. 17
Christakos, G. (1992). Random eld models in earth sciences. Academic Press Inc., San Diego. 18
Georgsen, F., Kolbjrnsen, O., and Lecomte, I. (2010a). A 3d ray-based pulse estimation for seismic inver-
sion of psdm data. In EAGE conference and exhibition. Extended Abstract. 28
Georgsen, F., Kolbjrnsen, O., and Lecomte, I. (2010b). The effect of illumination on source pulse estimation
for psdm seismic data. Technical Report SAND/10/10, Norwegian Computing Center. 28
Hampson, D. and Russell, B. (1990). AVO inversion: Theory and practice. pages 14561458. 14
Lecomte, I. (2008). Resolution and illumination analyses in psdm: A ray-based approach. The Leading Edge,
27:650663. 16
Lrtzer, G. J. M. and Berkhout, A. J. (1993). Linearized AVO inversion of multicomponent seismic data. In
Castagna, J. and Backus, M., editors, Offset-Dependent Reectivity Theory and practice of AVO analysis, pages
317332. Soc. Expl. Geophys. 14
Mosher, C. C., Keho, T. H., Weglein, A. B., and Foster, D. J. (1996). The impact of migration on AVO.
Geophysics, 61:16031615. 14
Pan, G. S., Young, C. Y., and Castagna, J. P. (1994). An integrated target-oriented prestack elastic waveform
inversion: Sensitivity, calibration, and application. Geophysics, 59:13921404. 14
Stolt, R. H. and Weglein, A. B. (1985). Migration and inversion of seismic data. Geophysics, 50:24582472. 15
White, R. E. (1984). Signal and noise estimation from seismic reection data using spectral coherence meth-
ods. Proceedings of the IEEE, 72:13401356. 25
CRAVA User Manual version 1.2 90
Index
Bayesian inversion, 14
CRAVA
implementation, 22
model le, 83
test suite, 86
userguide, 31
Crava model le
reference manual for elements, 48
estimation
mode, 48
FFT, 18
forward
mode, 48
inversion
basic, 31
mode, 48
JIRA, project management, 88
keyword
<3d-wavelet-tuning-factor>, 66
<actions>, 40, 48
<advanced-settings>, 63
<ai-file>, 77, 79
<ai>, 58
<allowed-parameter-values>, 42, 74
<angle-gather>, 67
<angle>, 34, 38, 53, 74, 82
<angular-correlation>, 67
<area-from-surface>, 33, 52
<ascii>, 57
<average-velocity>, 54
<background-trend-1d>, 62
<background-trend>, 59
<background>, 36, 43, 59, 77
<base-surface-file>, 72, 77, 81
<base-surface>, 34, 50
<blocked-logs>, 61
<blocked-wells>, 61
<bypass-coordinate-scaling>, 56, 68
<correlation-direction>, 80
<correlations>, 60
<crava>, 57
<debug-level>, 66
<density-constant>, 36, 78
<density-file>, 36, 78, 79
<density>, 38, 58, 73
<depth-file>, 35, 50, 51
<depth>, 55
<distance>, 72
<domain>, 39, 55
<dt>, 37, 73
<dts>, 37, 73
<earth-model>, 79
<elastic-parameters>, 39, 57
<energy-threshold>, 65
<error-file>, 38, 63
<estimate-background>, 49
<estimate-correlations>, 49
<estimate-local-noise>, 71
<estimate-scale>, 33, 69, 70
<estimate-shift>, 70
<estimate-wavelet-or-noise>, 49
<estimate-well-gradient-from-seismic>, 67
<estimation-interval>, 80
<estimation-range-x-direction>, 71
<estimation-range-y-direction>, 71
<estimation-settings>, 42, 49
<extra-grids>, 60
<extra-surfaces>, 62
<facies-likelihood>, 46, 60
<facies-probabilities-with-undef>, 44, 46,
60
<facies-probabilities>, 45, 46, 49, 59, 80
<facies>, 38, 73, 81
<fft-grid-padding>, 63
<file-name>, 34, 52, 67, 69, 70, 73
<file-output-prefix>, 38, 63
<filter-elastic-logs>, 38, 74
<format>, 39, 56, 61
<frequency-band>, 65
<global-wavelets>, 43, 62
<gradient-smoothing-range>, 67
<grid-output>, 38, 39, 55
<guard-zone>, 66
<high-cut-background-modelling>, 78
<high-cut-seismic-resolution>, 74
<high-cut>, 65
<il-end>, 34, 54
<il-start>, 34, 54
<il-step>, 34, 54
CRAVA User Manual version 1.2 91
<inline-crossline-numbers>, 33, 34, 53
<input-directory>, 38, 55
<interval-one-surface>, 51
<interval-two-surfaces>, 34, 50
<inversion-settings>, 40, 48
<io-settings>, 38, 55
<jason>, 61
<kriging-data-limit>, 66
<kriging-to-wells>, 40, 49
<lambdarho>, 58
<lame-lambda>, 58
<lame-mu>, 58
<lateral-correlation>, 78, 79
<length-x>, 34, 53
<length-y>, 34, 53
<local-noise-scaled>, 71
<local-noise>, 63
<local-wavelet>, 41, 69, 79
<local-wavelets>, 43, 62
<location-il>, 56, 68
<location-scaling-coefficient>, 57, 68
<location-x>, 56, 68
<location-xl>, 56, 68
<location-y>, 56, 68
<log-level>, 63
<log-names>, 72
<low-cut>, 65
<match-energies>, 71
<maximum-density>, 75
<maximum-deviation-angle>, 43, 76
<maximum-merge-distance>, 76
<maximum-offset>, 38, 76
<maximum-rank-correlation>, 76
<maximum-relative-thickness-difference>,
64
<maximum-shift>, 38, 76
<maximum-variance-density>, 76
<maximum-variance-vp>, 75
<maximum-variance-vs>, 75
<maximum-vp-vs-ratio>, 76
<maximum-vp>, 75
<maximum-vs>, 75
<maximum-wavelet-shift>, 65
<minimum-density>, 75
<minimum-horizontal-resolution>, 65
<minimum-relative-wavelet-amplitude>, 65
<minimum-sampling-density>, 65
<minimum-variance-density>, 76
<minimum-variance-vp>, 75
<minimum-variance-vs>, 75
<minimum-vp-vs-ratio>, 76
<minimum-vp>, 74
<minimum-vs>, 75
<mode>, 48
<murho>, 58
<name>, 81
<norsar>, 61, 62
<number-of-layers>, 34, 51
<number-of-simulations>, 40, 49
<offset-angle>, 32, 67
<optimize-location-to>, 38
<optimize-position>, 74
<original>, 59
<other-output>, 62
<other-parameters>, 39, 59
<output-directory>, 38, 55
<output-volume>, 33, 50
<parameter-correlation>, 37, 80
<poisson-ratio>, 58
<power>, 82
<prediction>, 40, 48
<prior-correlations>, 44, 62
<prior-model>, 36, 77
<prior-probabilities>, 81
<probability-cube>, 81
<probability>, 81
<processing-factor-file-name>, 70
<project-settings>, 49
<propagation-factor-file-name>, 70
<range>, 82
<reference-depth>, 54
<reference-point-x>, 34, 52
<reference-point-y>, 34, 53
<reference-surface>, 51
<reference-time-surface>, 55
<reflection-matrix>, 42, 66
<residuals>, 59
<ricker>, 33, 69
<rms-panel-mode>, 66
<rms>, 61
<rock-physics-distributions>, 46, 63
<sample-density-x>, 34, 53
<sample-density-y>, 34, 53
<sample-density>, 52
<scale-file>, 70
<scale>, 33, 69
<seed-file>, 49
<seed>, 40, 48
<segy-format>, 32, 56, 68
<segy-start-time>, 32, 67
<segy>, 57
<seismic-data>, 39, 59, 67
<seismic-quality-grid>, 46, 60
<sgri>, 57
<shift-file>, 69
<shift-to-interval-top>, 51
CRAVA User Manual version 1.2 92
<si-file>, 77, 79
<si>, 58
<sigma>, 72
<signal-to-noise-ratio>, 33, 71
<simulation>, 40, 48
<smooth-kriged-parameters>, 66
<snap-to-seismic-data>, 52, 53
<standard-format>, 56, 68
<start-time>, 68
<storm>, 57
<stretch-factor>, 71
<subrange>, 82
<survey>, 31, 67
<synthetic-residuals>, 59
<synthetic-vs-log>, 74
<synthetic>, 59
<task-file>, 38, 63
<temporal-correlation>, 37, 80
<thickness>, 52
<time-file>, 50
<time-gradient-settings>, 72
<time-to-depth-mapping-for-3d-wavelet>,
54
<time-to-depth-velocity>, 60
<time-value>, 34, 50, 51
<time>, 37, 55, 72
<top-directory>, 38, 55
<top-surface-file>, 72, 77, 81
<top-surface>, 34, 50
<type>, 41, 69
<uncertainty-level>, 81
<use-absolute-elastic-parameters>, 80
<use-for-background-trend>, 73
<use-for-facies-probabilities>, 73
<use-for-wavelet-estimation>, 73
<use-intermediate-disk-storage>, 64
<use-prediction>, 80
<use-vs>, 80
<utm-coordinates>, 33, 52
<variogram-keyword>, 82
<variogram-type>, 82
<velocity-field-from-inversion>, 36, 51
<velocity-field>, 35, 51, 78
<vp-constant>, 36, 78
<vp-file>, 36, 77, 79
<vp-vs-ratio-file>, 77, 79
<vp-vs-ratio-from-wells>, 64
<vp-vs-ratio>, 58, 64
<vp>, 37, 57, 72
<vs-constant>, 36, 78
<vs-file>, 36, 78, 79
<vs>, 37, 57, 73
<wavelet-3d>, 70
<wavelet-estimation-interval>, 71
<wavelet-output>, 43, 61
<wavelet-tapering-length>, 65
<wavelet>, 32, 69
<weight>, 38, 74
<well-data>, 37, 72
<well-move-data-interval>, 77
<well-output>, 38, 39, 60
<well-wavelets>, 43, 62
<well>, 38, 73
<wells>, 61
<white-noise-component>, 66
<x-fraction>, 64
<xl-end>, 34, 54
<xl-start>, 34, 54
<xl-step>, 34, 54
<y-fraction>, 64
<z-fraction>, 64
prior model
background, 22
reference manual, CRAVA model le elements, 48
release notes, 88
signal-to-noise ratio, 15
CRAVA User Manual version 1.2 93