L21 Course Summary

8/11/2019 L21 Course Summary

1/38

Seismic Imaging of subsurfacegeology

Course Summary


2/38

Seismic Imaging

Objective

Provide a framework to understand the underlying seismic model -how seismic data are affected by structure and stratigraphy and when

a seismic section is a good representation of geology

Course graduate should be able to

Communicate effectively with specialists in seismic acquisition,

processing, and interpretation

Understand how the earth responds to a seismic source and how

synthetic seismograms are generated in the computer.

Assess effects of earth filtering, data acquisition, and data

processing on seismic sections.

Appreciate Seismic data quality criteria

Resolution, Signal-to-noise ratio, Image integrity

Recognize whether appropriate technology has been applied to

your exploration project


3/38

Overview of Seismic Reflection Prospecting

Display

Seismic

Source

Seismic

signal is

generated

Seismic signal reflects

and transmits through

subsurfaceEarth

Filtering

DataAcquisition

Seismic signal isdetected and recorded

Data

Processing

(filtering)

Seismic signal is

enhanced

Imaging

(multi-dim.

Filtering)

Seismic signal is

positioned properly


4/38

Basic Seismology

Huygens principle is an extremely powerful wavefront tracking tool

By reconstructing wavefronts at closely spaced intervals, we can

propagate waves in uniform media

On prestack (CMP) gathers,

primary and multiple reflections are hyperbolicshape gets flatter (less moveout) with depth and increased

velocity

beyond critical angle, reflections turn into refractions

On poststack (zero-offset) sections,

Unmigrated events are mispositioned and look distorted;

diffractions abound, occurring at reflection discontinuities


5/38

Frequency domain, Linear systems,

& seismograms

Time and space signals can be expressed as sums of sinusoids

Time-space signals are represented by f-k Fourier transformsSignals must be sampled at least twice per period to avoid

aliasing

The output of a Linear Time Invariant System

has no frequency not present in input

is independent of order of operations

is convolution of system impulse response and input

can be obtained by multiplying Fourier transforms Seismic reflections are generated by convolving the earths

reflection coefficient train with a seismic wavelet.

The seismogram is further modified by ghosting (at source andreceiver) and attenuation


6/38

Fourier Transform Terminology

Bandwidth = frequency range where amplitude is

greater than 70% of maximum amplitude

Peak frequency = frequency at which amplitudespectrum has maximum value.

Broad band wavelets are those with large

bandwidths (typically more than 40 Hz for seismicdata).

Narrow band wavelets are those with smallbandwidths (typically less than 15 Hz for seismic

data).


7/38

-1.5

-1

-0.5

0

0.5

1

1.5

0 4 8 12 16 20 24 28

150 Hz Sine wave 100 Hz Sine wave

Aliasing: 4 ms sample rate (125 Hz Symmetry)


8/38

Importance of Fourier Transforms in

Acquisition and Processing

Traces can be filtered to remove low- and high- frequency noises.

Traces can be deconvolved to eliminate multiples or to enhanceresolution.

Resolution and dip control frequency needs, which controls

sample rate Field arrays filter surface waves and other noises in spatial domain

Space-time filters can eliminate surface waves, multiples, etc

Migration is a type of filtering, often performed in frequency

domain

Each frequency component can be treated independently

Signals and noises often have different frequencies

FFT (Fast Fourier Transforms) are computationally very fast

differential-equation solvers


9/38

Linear Systems Summary

The output of a Linear Time Invariant System has no frequency not present in input

is independent of order of operations

is convolution of system impulse response and input

can be obtained by multiplying Fourier transforms

Applications include filtering, synthetic (and real)

seismogram generation, attenuation and ghosting.

== dxxtvxhthtvty )()()(*)()(

)()()( fHfVfY =


10/38

Synthetic Seismogram Summary

Synthetic seismogram for vertically incident plane waveand horizontally layered media

Hang a seismic wavelet on each reflecting interface

Convolve the seismic wavelet w with the reflection coefficienttrain r

Multiply the FT ofw times the FT of r.

Ghosting plays a large role in wavelet shaping

Source and receiver each provides a ghost

Ghost is a 2-point (1,-1) filter separated by 2-way traveltimeto S or R depth. Notch in frequency domain at 0 and 1/ .

Attenuation increases with reflector depth

High frequencies are attenuated more than low frequencies Attenuation is constant for each fT (freq times 2-way travelT)


11/38

Seismic Data Quality Criteria(What does an interpreter want from seismic data?)

Reflection detection (signal-to-noise ratio,

overburden distortion)

Resolution (vertical: map thin beds, lateral: place faults,detect stratigraphic changes)

Fidelity (similarity of the seismic section to a geologiccross-section: accuracy of amplitudes, structural positions)


12/38

Important Concepts / Terms

Spread = arrangement of geophones relative to sources Off-end all receivers are on one side of the source Split-spread receivers are on both sides of the source

Group interval = receiver spacing= distance between geophone array centers

Channels = number of receiver stations recorded per shot

Shot interval = distance between shots

Bin is the collection of all traces with the same CMP (+/-) Fold is the number of traces in the CMP bin

CMP spacing = stack bin interval= 1/2 the group interval (usually)

Inline = direction of shooting

Crossline = perpendicular to Inline

3-D terms


13/38

Vertical Resolution: Key Points

Resolution is the ability to distinguish distinct events Short-duration seismic wavelets are required

Broad bandwidth, zero-phase wavelets are best

Seismic bandwidth is approximately = center frequency

Potential resolution

Bandwidth can be increased by deconvolution Frequencies to be included must have adequate S/N

Phase coherence or Wiener spectra determine S/N

Deconvolution must

Increase the bandwidth

and / or Align frequency components to be in phase


14/38

Stratigraphic exploration: Amplitudes

Thin Beds

Reflection shape changes little with bed thickness

Reflection amplitude is proportional to bed thickness Amplitudes

Amplitudes convey useful information about rock properties

Statistical AGC makes data useful for structural

interpretation

Long-gate AGC better than short gate for amplitude fidelity

Controlled amplitude - deterministic - most accurate

approach but difficult to remove all amplitude contaminants

S i i I i


15/38

Seismic Imaging

Vertical Resolution

Good vertical resolution is needed to isochron thin beds

Resolution requires broad-band, zero-phase seismic data

Statistical deconvolution broadens the bandwidth & simplifiesphase, subject to acceptable signal-to-noise ratio

Deterministic approach is an excellent alternative (when source

and instrument characteristics are available.)Lateral Resolution

Good lateral resolution is needed to map small structural features

faults, pinnacle reefs, contorted beds

Fresnel zone smearing causes poor lateral resolution at depth

Migration can improve lateral resolution significantly

Prestack migration is especially helpful in improving lateral

resolution


16/38

Seismic Visibility of a bed


17/38

Vertical Resolution / Gain Enhancement Deconvolution / AGC

Applies a different filter / gain to each trace

Assumes random reflection amplitude & depth Statistical correction

Controlled Amplitude & Phase (CAP)

Requires coordinated acquisition & processing Applies same filter / gain to all traces

Assumes all relevant parameters are known

Deterministic corrections

Use CAP whenever possible

Requires special acquisition & processing

Requires fairly high-quality (good S/N) data


18/38

Rock Physics Summary

Seismic response depends upon velocityand density

Velocity and density depend upon rock

properties

lithology, porosity, fluid content, age, depth of

burial, pore pressure, heat flow Seismic amplitude variation with offset also

depends upon shear-wave velocity


19/38

Summary:

Stratigraphic modeling / inversion

Used to extend well-log information away

from the well

Matching synthetic seismic data to real datapermits checking geologic hypotheses

Result is an estimate of short-periodvelocities

used to infer rock properties away from wells


20/38

Velocity Summary

Velocities are the most important factor in seismic explorationLong period estimates

Obtained from traveltime information

Processing uses: NMO Stack, MigrationInterpretation uses: Well-seismic ties, depth estimation,

overpressure prediction

Short period estimatesObtained from reflection waveforms and amplitudes

Interpretation uses: rock property prediction away from

wells


21/38

21

Velocities and CMP stack

Velocities are at the heart of seismic analysis

NMO velocities can be used to estimate interval

velocities and aid in time-to-depth conversion

NMO velocity errors cause stack to act as a filter -

eliminating high frequencies

Fold is extremely important in enhancing S / N

Offset / azimuth distribution also very important

Shooting geometries to achieve CMP stack often dictated by

equipment availability / field terrain


22/38

Basic Acquisition and Processing

Objectives of acquisition and processing:

reflection identification, resolution, and fidelity

CMP Shooting forms the basis of acquisition and processing 3 - D fold = (Channels per streamer / 2) * (Group Int / eff Shot Int)

Acquisition, the most expensive step in seismic exploration, tries

for the best data quality at a reasonable cost

Illuminate target with sufficient source energy

Minimize recorded noise

Basic Processing comprises

Editing and sorting into CMP order

NMO velocity analysis and NMO application CMP stack


23/38

Basic 2-D Processing Stream

Filtering, corrections

Sorting, Labeling, Editing

CMP Stack

Filtering, migration

Velocity

analysis

Pre-

Stack

Post-

Stack


24/38

Signal-to-Noise Ratio

CMP stack enhances S / (random noise) by n, where n =fold

enhances S / (coherent noise) with stack array

f-k filters suppress dipping noises

Multiples are most significant marine data quality

problem In very shallow water, deconvolution is effective

In very deep water (where sufficient differential

moveout is present), velocity filtering techniques areeffective : f-k filters, weighted stack, radon filters

Surface-related multiple suppression is a recent

development for complex geometries


25/38

Seismic Arrays and CMP Stack

Virtually all exploration seismic data are detected

by receiver arrays The alternative, a Multi-component detector, is in itsinfancy

Seismic Arrays are effective at reducing short-wavelength coherent noises in the inline direction

Long arrays attenuate reflection signals

Field arrays are anti-alias filters

Spatial frequencies that would alias are attenuated


26/38

Signal-to-noise enhancement

Full-fold Stack Array reduces surface-wave noiseswithout significantly attenuating reflections.

Full-fold data acquisition used only in areas with severe

noise problems, due to cost.

Less than full-fold data acquisition produces stack array

response that attenuates surface-wave noise, but to a

lesser extent.


27/38

Data Processing Terminology

Basic data processing consists of

enhancing signal-to- noise ratio

accounting for traveltime variations (in overburden and

with offset)

compensating for geologic structure

High-end processing is needed

when basic data processing does not meet business needs to compensate for complex structure / overburden

to attenuate complex noise fields


28/38

Advanced Data Processing Summary

Data processing streams must be tailored to source of

data-quality degradation Filtering is often useful in eliminating additive

noises

Efficacy of multiple suppression technique depends

on water depth

Long-period static correction reduces structuraldistortion

Short-period static correction enhances signal-to-

noise ratio


29/38

Addressing Seismic Data Quality

Reflection detection (signal-to-noise ratio, overburdendistortion)

1-D and 2-D filters aimed at attenuating specific noises CMP Stack

Short-period static and dynamic corrections

Resolution

Vertical: Deconvolution / controlled phase processing

Lateral: Migration Fidelity

Amplitudes: Controlled amplitude processing

Structure: Long-period statics, Velocity estimation, Timemigration, Time-depth conversion, Depth migration


30/38

Structural Modeling summary

Structural modeling provides insight into the

relationship between a complex geologic structure

and its seismic response and is often useful in

interpreting complex seismic data.

Raytrace modeling is inexpensive and providesinsight into the seismic-geology relationship.

In recent years, wave equation modeling has

become much more affordable.

Migration is the inverse to structural seismic

modeling.

Marmousi Model


31/38

Marmousi Model

Velocity

Density

Versteeg

TLE, 2004


32/38

Migration Summary - I

Migration makes a seismic section look more like a geologic cross-

section

Flat horizon remains unchanged, if no velocity anomaly above it

Dipping horizon becomes steeper, shallower, and moves laterally

up dip

Synclines become broader and bow ties are eliminated

Anticlines become narrower

Diffractions collapse to points

Migration operationCorrect velocities are key to successful migration

Noise spikes and edges cause migration smiles

Adequate aperture needed to capture dipping eventsA properly migrated section has neither diffractions nor smiles

Migration Summary II


33/38

Migration Summary - II

Depth migration is necessary when velocity varies rapidly(laterally) but is more expensive than time migration

Accurate velocity-depth model is essential

Kirchhoff migration used almost exclusively (adequateaperture is important)

Prestack migration overcomes limitations of stack but is more

expensive than poststack migration

Depth point smear,, conflicting dips

NMO combined with DMO, stack and poststack migration iseffective and less expensive than full prestack time migration

3-D migration does not require all energy to come from directly

below the seismic line

eliminates sideswipe

moves energy in both inline and crossline directionsPrestack 3-D Depth Migration combines all of the above (costly)

Survey Surface Coverage


34/38

Survey Surface Coverage

Shots occupy a distance = prospect length + ApertureLeft + ApertureRight + 1/2 spread

Line length = positions occupied by either sources of receivers = prospect length +ApertureLeft + ApertureRight + 1 1/2 spread

ProspectSeismic line

Aperture Aperture

Full-fold

Shot point 1

Shooting direction

(off-end shooting) Full-fold buildup(1/2 spread)

Full-fold buildup(1/2 spread)

Last shot

point

streamer


35/38

Summary of Migration algorithms

Correct velocities are key to successful migration

Noise spikes and edges cause migration smiles

Adequate aperture needed to capture dipping events

A plethora of migration algorithms are available

Historically, Kirchhoff migration has been most

widely used

Wave-equation methods have now become

affordable

C i h h h i (i li )


36/38

Cross-sections through the reservoir (inlines)

Inline

From well data, AVO analysis, & seismic inversion, obtainpetrophysical information to populate sublayers

Pl i 3 D S


37/38

Planning 3-D Surveys

Pre-design stage

Determine objective

Assemble required informationDesign Stage

Determine resolution

Determine aperture

Determine sampling

Estimate Effort level (fold)

Post-design stage

Feedback results of survey

Iterate design process

Interpreter

Designer

Interpreter

Designer

Processor


38/38

Survey Design: Summary

Important objectives: small vertical and lateral resolution high signal-to-noise ratio

small acquisition footprint

low cost and fast acquisition time

L21 Course Summary

Documents