Basic interpretation

1
Basic interpretation
2
Basic steps

• Examine data
• problems with processing? Misties?
• Be familiar with regional geology and background
  but try not to make assumptions (compression in
  Gulf Coast? Usually not, except at toe of growth
  faults)
• Goal
• Pick reflections (horizons)
• Peak, trough, zero-crossing?
• Post to map and cross-sections
• Often need to pick faults first
• Start with big ones
• Know style of faulting in area
• Map in loops and make sure that all horizons
  and faults are consistent (tie)

3
Faults

• Termination and offset of reflections
• Different dips across fault
• Shadow zone below fault
• Diffractions on unmigrated data
• Fault itself usually does not make a reflection
• - angle too steep for most processing
• - velocities and reflection coefficients vary
  quickly
• - fault zone rather than single break
• Fault appear to curve with depth due to velocity
  (even if straight)
• Often appear above underlying features (but not
  always e.g. growth faults)
• Faults do not have to link up
• - bedding plane slip or below resolution
• - causes space problems in models

4

• Procedure
• Preview the data set using various displays
• Processing problems
• Regional geology
• Structural style
• use volume tools if available
• Develop fault structure
• Map horizons (or sequences)
• Transfer to model

5
Display

• Wiggle trace
• Variable intensity color (or grayscale) scaled to
  amplitude (or voxels)
• balanced appearance
• no overlap or clipping
• no mislocation of high amplitudes
• higher dynamic range
• Single-color (e.g. grayscale) good for structure
• Double color good for stratigraphy
• Blue (positive)-white-red (negative)
• More colors for higher detail

Wiggle trace
6
Time slice
7
stratigraphy
8
Other attributes are available
Coherence similarity between traces good for
faults Instantaneous phase, frequency Dip
magnitude Edge detection Multiple attribute
combinations
Coherence (top) and amplitude (bottom) of salt
dome
9
Preview the data
Use volume and opacity for quick view of surfaces
without interpretation Spot amplitude anomalies
(gas?) Use time slices Get a sense of overall
structure
From Dorn (1998)
10
2D interpretation/volume interpretation

• 2D pick faults and horizons on 2D images
• Integrate all picks to develop a 3D view
• Same as done on paper sections
• Volume use transparency and surface perspective
  to see horizons and structure

11
Works well with good data Usually have to fall
back to 2D picking eventually
12
Modern seismic interpretation programs

• Landmark, Geoquest are the majors a variety of
  second tier software.
• Combination of interpretation, visualization,
  GIS, and modeling
• Workstation or PC (both expensive)
• Read in processed seismic data and display
• Allow picking of horizons and faults
• Store picks in relational database (along with
  well data such as sonic logs)
• Create maps directly from picked data
• Analyze data (isopach maps, flatten horizons,
  etc)
• Linked to modeling programs (e.g. GOCAD) to
  create internally consistent models
• Lots of options.

13
Data loading and organization
Most load in SEGY (trace data) as well as formats
from other programs Require detailed geodetic
information (geoid and ellipsoid) so that maps
are accurate within meters (for later
drilling) Data is organized into projects with
all data (seismic, picks, well, etc) from a
single area Seismic data is in surveys may
have different vintages (years) Can be difficult
to compare different vintage data due to
processing and acquisition differences Often have
misties slightly different depths for the same
reflector
14
Surveys

• 2D data lies along a line
• 3D data in two directions
• Inline data parallel to direction it was shot
• Crossline perpendicular to direction it was
  shot
• Spacing between inlines is usually less than
  spacing between crosslines
• Not usually parallel to compass directions
• 4D usually refers to 3D data over the same area
  at different times used to monitor production,
  steam flooding, etc (Im not sure what 2D data at
  different times is called)

15
Faults and horizons

• Basic geometric objects used in interpretation
• Horizons follow distinct reflectors or sequence
  boundaries
• Geometry assumed fairly simple (cannot overlap 1
  horizon on itself)
• Each horizon stored as a separate file
• Usually use snapping to even out picks
• Can interpolate picks to form a surface
• Can flatten on a given horizon to remove effects
  of structure
• Subtract two horizons to create an isopach map
• Create maps of amplitude changes along a horizon
• Faults
• Can join each other
• Usually associated into groups of faults
• Cannot use snapping
• Can interpolate
• Can edit, modify and delete all picks easily
  (sometimes too easily)

16
horizons

• Usually, an amplitude peak that corresponds to a
  change in acoustic impedance (i.e. lithology)
• Can be a trough or even zero-crossing.
• The ideal horizon is a zero-phase wavelet
  centered on the change.
• Seismic sources cannot produce a zero-phase
  wavelet but are often minimum phase.
• The earth filters out high frequencies also
• Can adjust with processing.

T0
Zero-phase
Minimum phase
17
Automated picking
Snapping Max, min or 0 amplitude within a set
distance of the pick Automatically jumps to the
nearest high amplitude Must be close (otherwise
jumps to wrong peak) Autopick Pick one spot on
horizon Snaps to peak then picks nearest peak on
next trace Works great in good data with no
faults 3D version selects all voxels with a
certain amplitude a cloud
From Dorn (1998)
18
Resolution
The minimum separation between two features such
that we can tell that there are two features
rather than one. Horizontal resolution how far
apart two features on single interface must be to
show as separate features Vertical resolution
how far apart two interfaces must be to show as
separate interfaces Depends on the size of
seismic wavelets which have a limited frequency
range and therefore width (the limit is the
Dirac delta function which contains all
frequencies)
19
Horizontal resolution
Reflections occurs over a area and not just at
point due to the wave nature. Fresnel zone is the
area from which the phase of the reflection from
a point source (curved wavefront) differs by less
than a half-cycle First Fresnel zone is where V
is average velocity, l frequency, and t is the
time.
Strictly speaking, this is only true for
unmigrated data. Migration collapses
diffractions, so migrated data should have better
resolution.
20
Vertical resolution

• Depends on wavelength the usual criteria is ¼ l
  (wavelength)
• How thick is a thin bed?
• Less than ¼ l top and bottom appear as a single
  reflection
• A very thin bed with a high reflection
  coefficient will still be visible
• Decreases with depth

21
Resolution of fault throw
22
Tuning how thin is a thin bed?

• As a bed gets thinner and approaches l/4, the
  amplitude will very due to destructive and
  constructive interference from the top and bottom
  reflections
• The reflections will also be shifted in time
  giving an inaccurate thickness
• Critical for stratigraphy
• Another reason why synthetic seismograms can fail

tuning
From Sheriff different velocity models
23
Phase and polarity
Zero-phase a reflection produces a wavelet that
is symmetric on each side of the reflecting
surface. Ideally, modern seismic reflection data
is processed so that this is true (but doesnt
always work). Minimum-phase An alternate
representation. Wavelets can be positive or
negative. One way to check with marine data is
to look at the water bottom reflection, as it
always represents an increase in acoustic
impedance. Reflections which have the same
polarity as the water bottom reflection represent
positive changes in acoustic impedance.
Minimum phase
Zero phase
24
Increasing resolution - VSP

• Vertical Seismic Profile
• Drop geophones down a well and put source at
  surface
• Only goes once through top layers less
  attenuation of high frequencies
• Closer to reflections
• Higher resolution

normal
Downgoing wavefield
VSP
Vsp data superimposed on normal
data (Baker-Hughes)
Upgoing wavefield
25
Wait a minute

• If we have a series of beds that are spaced
  evenly at some fraction of a wavelength, then the
  small reflection can constructively sum to form a
  large reflection.
• Not all large reflections are due to a single
  strong boundary
• In fact, not all small reflections are due to a
  single small boundary
• If we look at a well log, we see all kinds of
  variations.
• This was not fully realized until the 1970s

26
What is a reflection, anyway?

• Last month we said it was an acoustic impedance
  boundary true.
• That ignored wavelet effects destructive and
  constructive interference.
• A seismic wavelet used in reflection seismology
  is on the order of 10 to maybe 100s of meters
  long
• Sedimentary rocks often have evenly spaced layers
  with thicknesses on that order
• In 1977 Peter Vail (a geologist) claims that a
  reflection is a chronostratigraphic surface
• This is the basis of seismic sequence
  stratigraphy.
• It made geophysicists uneasy, but seismic
  sequence stratigraphy works.

27
Chronostratigraphic surface

• A surface deposited at the same geologic time
• Basically a bedding plane
• A single surface might divide sandstone and shale
  in one place and shale and limestone in another
• But it would be all one reflection

A regressive sequence
limestone
chert
shale
Unconformity but also a reflector
ss
shale
limestone
28
Seismic and stratigraphy

• Seismic-sequence analysis
• Use time-depositional units
• Map unconformities
• Seismic facies analysis
• Estimate depositional environment from reflector
  configuration
• Not an exact science
• Reflection-character analysis
• In-depth study of a single reflector
• Use synthetic seismograms
• Direct detection of hydrocarbon
• Ultimate goal of petroleum exploration
• Currently possible in some cases

29
3D fault picking

• Strategy
• pick all faults as generic and assign later
• depends on structure, number, and complexity
• Break data into grid with similar spacing
• set roll increment to say, 8 or 16 (use multiple
  of 2 to make filling in easier later)
• decimate traces to make spacing even in inline
  and crossline directions
• Do not agonize over individual faults at first
  pick them and revise later
• pick a series of parallel lines and then verify
  with perpendicular or time slices
• Look at throw of faults (throw increase with
  depth for growth faults)
• Look at dip of fault blocks

30
Map Colors show fault picks on individual lines
Crossline circles are postings
Inline faults picked here
Time slice circles are postings
31
Horizon picking

• Have major faults picked already
• Usually pick initially on 1 crossline or inline
• Then switch to perpendicular direction and
  starting from the single visible point
• Go in loops to ensure all ties are correct
• Horizons end at faults can mark up or down for
  maps
• Snap to peak amplitude
• Can use autopick inside closed polygons

32
Structural style

• Basement involved
• - extension
• - compression
• - wrench
• Basement-detached
• - thin-skin thrust
• - growth faults
• - salt and shale tectonics

33
(No Transcript)