Vous êtes sur la page 1sur 8

gain recovery (energy loss: see class notes page 1.

1-27) correct for the effects of spherical divergence (1/r loss of amplitude) multiply data by r (distance traveled) in practice, velocity usually increases with depth, causing additional geometrical spreading for 1D earth and near-normal incidence
2 gain = r v rms t

where vrms is the root-mean-square velocity from the surface to time t determined from the data (described in next section of class notes) often simplified by multiplying data by power of t

gain = t
most often, = 2 produces good results

correct for the effects of attenuation multiply data by an exponential gain

gain = e t

>1

attenuation is greater for higher frequencies deconvolution can attempt to recover the higher frequencies

Hole: GEOS 4174

4.1-8

Reflection Data Processing: Prestack Signal Processing

gain recovery narrow-bandpass-filtered shot gathers before (left) and after (right) spherical divergence correction

Yilmaz 2001

higher frequencies are more attenuated, and are not recovered by gain correction

Hole: GEOS 4174

4.1-9

Reflection Data Processing: Prestack Signal Processing

Caspian data set: shot gathers after spherical divergence (t2) correction

Yilmaz 2001

later primary reflections are much stronger later noise is also much stronger
Hole: GEOS 4174 4.1-10 Reflection Data Processing: Prestack Signal Processing

geometry assignment observers notes connect trace number to shot site number and receiver site number surveyors notes connect shot and receiver sites to geographic location enter these meta-data into the processing stream shot number, shot x, shot y, shot z (at surface), shot depth receiver number, receiver x, receiver y, receiver z computed shot-receiver offset, azimuth computed midpoint number, midpoint x, midpoint y, midpoint z this is the processing step most prone to human error elevation statics if seismic line is not on flat topography, seismic traces from different shot-receiver pairs will not line up at a deep reflector flatten traces to a new datum elevation using a static time correction:

t = t SD + t RD =

E is elevation at Source, Receiver, or new Datum level D is depth of shot in borehole V is determined from uphole time recorded at geophone above shot
or using direct arrivals in shot gathers
modified from Reynolds 1997

E S E D DS E R E D + V V

floating datum: if elevation variation is significant, use a datum that is a very smooth version of the true topography
Hole: GEOS 4174 4.1-11 Reflection Data Processing: Prestack Signal Processing

field statics usually there is a very slow weathering layer in the near-surface (the low-velocity layer) varies along the line, so needs to be included in the statics flatten traces to a new datum elevation using a static time correction:

E S E D DS E R E D DR t = + + t uh Vb Vb t uh is uphole time of the shot co-located at the receiver position Vb is determined from refracted arrivals recorded in shot gathers
R
R

Yilmaz 2001

refraction statics often drilled shots are not used (no uphole times) sometimes there is complex or multi-layered near-surface velocity structure => determine 2D or 3D near-surface velocity structure directly from first arrivals in shot gathers (refraction data analysis for velocity structure will be covered later in the course)

t = t SD + t RD
the two times are determined directly from the 2D or 3D velocity model

Hole: GEOS 4174

4.1-12

Reflection Data Processing: Prestack Signal Processing

deconvolution (method discussed in class notes section 3.3) predictive deconvolution to remove multiples spiking or predictive deconvolution to shorten the source wavelet bandpass filter (method discussed in class notes section 3.2) low-frequency noise: e.g., ground roll high-frequency noise: e.g., wind noise, noise from deconvolution trace balancing recording conditions and/or single-trace deconvolution can apply unequal amplitude scaling to different traces apply gain to each trace so that its amplitude is consistent with neighbouring traces

gain =

constant 1 xi 2 N i=1
N

constant gain = N 1 xi N i=1

where the time window for the summation is large and contains high-quality data the scaling on the left is the root-mean-square (rms); the right is mean-absolute fk filter (method discussed in class notes section 3.2) coherent noise: ground roll, water wave, out-of-plane scattering statics corrections should be applied first so that coherent events line up improved results are sometimes achieved by fk filtering of both common-shot and common-receiver gathers (requires a data sort, then re-sort)
Hole: GEOS 4174 4.1-13 Reflection Data Processing: Prestack Signal Processing

Caspian data set: shot gathers after spiking deconvolution

Yilmaz 2001

wavelet is compressed reverberation is reduced

Hole: GEOS 4174

4.1-14

Reflection Data Processing: Prestack Signal Processing

Caspian data set: shot gathers after wide bandpass filter and trace balancing

Yilmaz 2001

removed low and high-frequency noise (primarily from deconvolution) trace energy is consistent

Hole: GEOS 4174

4.1-15

Reflection Data Processing: Prestack Signal Processing

Vous aimerez peut-être aussi