Part I

Static corrections—a review

By DAVE MARSDEN
Amoco (UK) Exploration Company
London, England

I n many onshore exploration areas, the land surface is cov- upper layer can cause a dramatic deterioration in the quality
ered with a relatively thin layer of material of low seismic of land seismic data (Figure 1) if we do not acknowledge the
velocity. Commonly known to geophysicists as the weath- problem and take appropriate action during data acquisition
ered layer, it is generally related to aerated material above the and processing.
water table or to geologically recent unconsolidated sedi- There are many concerns and issues associated with the
ments on a substratum of harder consolidated rocks. This near surface but, in this three-part tutorial series, I shall be
seismic layer, despite the geophysicist’s terminology, ap- concerned with the problems of velocity and thickness vari-
pears to have very little to do with the geologic weathered ations in the low-velocity, near-surface layers-or, more
layer. However, variations in the physical properties of this particularly, with their measurement and compensation of the
data for their presence. Part I deals with the variability of the
near-surface velocity distributions and how it affects the
seismic data. Also considered are the types of data available
from which the near-surface velocity distribution may be
determined. Concluding the first part is a discussion of the
datum, field statics and drift statics. Next month, Part II will
consider refraction statics, and in March Part III will focus on
residual statics, statics and velocity analysis, plus some guide-
lines for choosing a method.

P roblems caused by the near-surface low-velocity layer

have been known for over half a century. Some of the earliest
papers in geophysical prospecting were concerned with at-
tempts to determine their thickness and velocity, or compens-
ate those early seismic records for the time delays caused by
the low-velocity layer. In predigital days, field statics and
refraction statics were thought to be the complete statics
solution; then, in the wave of the success of residual statics
programs (first developed in the 1970s), it was felt that
statistical methods alone were the answer. However, the
consensus today within the exploration industry is that each
method has its own place in adding to the complete statics
solution.
Despite the many technologies that deal with different
aspects of the near surface, related issues are still with us. Two
of the most difficult, and most often cited, problems are:

l Need for more accurate near-surface velocity models

l Need for models of the near surface to allow adequate
acquisition design
Figure 1. (a) Part of a seismic line processed without static
corrections. (b) Same data processed with static correc- Before we can model the near surface, we need to measure
tions. Note how the resolution and continuity of events are it so that our models bear some relationship to reality. What
improved in the latter (from Residual statics analysis as a we usually measure are the velocities and bed thicknesses, or
general linearproblem by R.A. Wiggins et al., GEOPHYSICS some kind of averages of these since we tend to ignore very
1976). thin beds even though their presence or absence can affect the

THE LEADING EDGE JANUARY 1993 43

 apparent phase of the data (see A simple approach to high
resolution profiling for coal by A. Ziolkowski and WE.
Lerwill, Geophysical Prospecting 1979).
Another aspect of the near surface which we shall not
discuss is the question of ground roll, even though the fre-
quency and velocity of these surface waves are governed by
the thickness and elastic properties of the near surface. Ac-
quisition parameters are usually selected so that the nuisance
value of ground roll is minimized, although, as we shall see,
the way we choose to overcome the problem can lead to a loss
of temporal resolution. Today, we are acquiring seismic data
in more rugged areas and areas that were formerly “no data” Figure 2. Some of the frequently encountered near-sur-
areas, and pushing existing technologies to the limit of their face conditions which, if not adequately modeled, result
usefulness in the process. However, we still experience prob- in errors in the computed static corrections and a de-
lems in more normal areas, some with a long history of graded seismic image.
seismic activity. “Why is this?” you might well ask. The
principal reasons are:

• Need for higher resolution data demands better static

corrections among other things
• Failure to consider the near surface and its problems at
the survey planning stage can result in not acquiring the
necessary data to address the issue
• Ignorance of the problem
• Lack of awareness of the available technologies
• Cost cutting

Static corrections are most important in the processing of

land data for they lead to improved quality in subsequent
processing steps which, in turn, impact the integrity, quality,
and resolution of the imaged section. Errors in the static
correction lead to a loss of seismic resolution, both temporal
and spatial, and a less-than-optimum interpretation of the
seismic data set. Also, if static corrections are not properly
derived, then a whole catalog of problems can beset the
interpreter (such as lines with variable datums, seismic events
which mistie at intersections, false structural anomalies re-
maining in the data, false events being created out of noise,
and lastly the data quality not being optimized). Therefore, a
good statics solution is desirable for two reasons: to obtain
the correct structural interpretation and to obtain a high-res- Figure 3. (a) A depth model. (b) The model’s seismic time
olution section which can be used for stratigraphic interpre- response, illustrating the fundamental issues of the statics
tation. It should be stressed that either of these criteria can be problem. Changes in the elevation and thickness of the
met without satisfying the other by application of one or near-surface low-velocity layer produce time structures
another of the different statics technologies that are available; on reflections from flat reflectors. Lateral variations in
however, it is most desirable to satisfy both. the interval velocity of the near surface have similar
A literature search has yielded only one tutorial on the effects.
subject of static corrections-a 1989 article by Brian Russell
which appeared in the CSEG Recorder, a publication not
widely available. The purpose of this three-part series, then,
is to attempt to correct this lack of up-to-date, easily accessi-
ble information on this important subject. The articles will
consider the problem of velocity and thickness variations in
the low-velocity, near-surface layers; discuss the types of data
available for determining static corrections; analyze altena-
tive approaches; and review available technologies.

N ear-surface concepts. In many exploration areas, the

surface is covered with a relatively thin and uniform low-ve-
locity layer, but frequently we know that this is not the case.
Some of the near-surface conditions which are frequently met Figure 4. Near-surface model with a seismic raypath
are illustrated in Figure 2. They include, but are not limited being shown between source and receiver.

44 THE LEADING EDGE JANUARY 1993

to, elevation changes, sand dunes, other eolian deposits, average velocity between the datum and the reflector.
buried river channels, buried glacial scours, permafrost, evap During processing, each of the above effects usually un-
orites, variable water table, leached zones, volcanics, peat dergoes one or two corrections at a time, until the seismic data
deposits, and coal seams. All present potential problems. The provide a quality image of the subsurface. With conventional
presence of low- or high-velocity layers by themselves do not multifold data, a number of traces are added together in such
pose problems. The problems occur because of the variability a way that the summing, or stacking, enhances primary re-
in both the thickness and velocity of the near-surface layers flections at the expense of noise or unwanted signal. Correc-
and our ability to adequately define those variations or com- tions applied to the seismic traces so that the data can be
pensate for them. properly stacked are of two types, static and dynamic. Static
The simple model shown in Figure 3 illustrates the essence corrections involve a constant time shift to the data traces
of the near-surface problem. The depth model has a variable whereas dynamic corrections involve time variable shifts.
overburden thickness due to elevation changes and other Corrections made to each seismic trace for elevation effects
effects and its interval velocity is assumed constant. The (elevation static) and near-surface low-velocity effects (weath
attitude of the seismic reflections clearly do not represent the ering static) by conceptually moving the shots (shot static)
structural attitude of the reflectors in the depth model. Similar and receivers (the receiver static) to a common reference
effects could be produced by holding the overburden thick- surface (the datum plane) are greatly simplified if it is as-
ness constant and varying its interval velocity. Where the sumed that energy travels vertically in the interval above the
overburden is thicker (or of lower interval velocity), a seismic datum plane.
wavelet takes longer to travel through the layer and con- In conventional seismic surveys, shot-receiver separation
versely where it is thinner (or of higher interval velocity), a is usually less than or approximately equal to the depth to
seismic wavelet requires less time to traverse the layer. reflector so, with reference to Figure 4, we see that in practice
Seismic recording involves a source and receiver, usually the raypaths in the near surface tend to be almost vertical.
many receivers, separated by some offset distance. The Such an assumption results in static, rather than dynamic,
raypath for a single reflection on a seismic recording is shown corrections with very little residual error in most cases. This
in Figure 4. From this and Figure 3, we can see that the assumption allows all traces from a common shot location to
traveltime of a wavelet along the raypath is influenced by the receive the same shot static and all traces from a common
surface elevations of the geophone and shotpoint, by the receiver location to receive the same receiver static. The static
velocity and thickness of the near-surface layers above the corrections then are said to be surface-consistent. The as-
datum, by the depth and dip of the reflector itself, by the sumption of vertically traveling energy is most closely ap-
distance separating the source and receiver, and lastly by the proximated for small offsets, for small elevation changes, for

Figure 5. Effect of statics on the stacked trace. (a) Syn-

thetic 12-trace gather, after NMO corrections, but con-
taining uncorrected static anomlies. (b) Stack of the 12
traces repeated so the degraded reflection is clearly seen. Figure 6. A small static shift between two traces acts as a
(c) Stack of the 12 traces after statics correction, once high-cut filter. The frequency attenuation, when two
again repeated so that the reflection is clearly seen (from traces are summed (or the output from two geophones in
Static corrections-a tutorial by B.H. Russell, CSEG Re- an array are summed), is a function of the static shift
corder 1989). between them.

THE LEADING EDGE JANUARY 1993 45

 Figure 7; The field geometries used in conventional seismic profiling do not sample the very near-surface layers
adequately enough to permit a sufficiently detailed velocity model to be determined from the directly arriving seismic
energy.

small elevation differences between topographic surfaces and

datum, and in those situations where there is a relatively thin
veneer of low-velocity material at the surface. The assump-
tion of vertical raypaths breaks down for data recorded in the
most rugged terrains where elevation changes along the line
are large with respect to the distance between geophone
groups, for large differences between topographic surface and
datum, where the velocity at the surface is relatively high and,
lastly, for offsets which are long with respect to the depths to
the reflectors.
Static anomalies cause the primary events to be misaligned
and so the stack process, rather than enhancing the primary
signal, destroys or corrupts it. Figure 5 illustrates the effect
that static anomalies have on the stacked trace. In Figure 5a,
a theoretical 12-trace gather is shown, prior to stacking, which
contains some small static anomalies. These anomalies cause
the primary event to be misaligned across the gather so that
when the traces are stacked, a double event of low amplitude
is produced (Figure 5b) rather than the single high amplitude
event (Figure SC) which would be the desirable outcome of
the stacking process. In real data, these short-period static
anomalies act as high-cut filters on the amplitude spectrum
and also cause phase distortion of the wavelet shape.
Figure 6 relates the magnitude of the static shift between
two summed traces to the attenuation of different frequencies.
Consider, for example, traces 1 and 2 in Figure 5a. The
seismic wavelet has a dominant frequency of about 70 Hz,
and there is a time shift of about 5 ms between the two traces.
Summing just these two traces without any static correction
results, according to Figure 6, in the rejection of the 100 Hz
data from the amplitude spectrum of the summed trace (i.e.,
there is a notch at 100 Hz). The data close to the dominant
frequency (about 70 Hz) are reduced by about 6 dB-i.e., its
amplitude will be half of what it would be with a static
correction applied, while the signal at lower frequencies is
little affected. We thus see that the higher frequencies are
more severely attenuated than the lower frequencies. This
Figure 8. A vibroseis record where the first breaks (onset helps explain why we so rarely see frequencies above about
of seismic energy) cannot be distinguished. In such areas, 70 Hz in land seismic data when the commonly used sources
it is necessary to use an additional source of information are capable of producing frequencies in excess of 100 HZ-
if the near-surface velocity model heeds to be derived for static shifts of 4-5 ms are easily overlooked in processing
(from Seismic Data Processing by 0. Yilmai, SEG, 1987). and easily produced within a typical receiver group length by

46 THE LEADING EDGE JANUARY 1993

Figure 9. The field geometry of a low-velocity layer survey, showing typical geophone and shot positions, permits the
velocity model of the near surface to be determined by refraction analysis of the field records. Compare the scale of the
geometry to that shown in Figure 7.

variations within the soil layer.

The above serves to illustrate that as we strive for higher
resolution data, for detailed stratigraphic exploration, close
attention needs to be given to static corrections which are
conceptually very simple. We derive the thickness and veloc-
ity of the near-surface layer and, given the elevations of all
shot and receiver locations, we compute and apply the static
corrections to each trace in a surface consistent manner. The
commonly available methods to do this assume that the near
surface consists of one or two low-velocity layers. This
limitation does not pose a serious problem since most of these
techniques can be used iteratively (i.e., the first layer is
analyzed and a correction made before considering the sec-
ond layer). However, if the near-surface model is inade-
quately defined (and it is usually oversimplified), there will
be trade-offs between the velocities and thicknesses of the
layers in the model which result in errors in the computed
static corrections and a less-than-optimal seismic image.

Data sources. How well we define the surface model or Figure 10. (a) Acquisition geometry of an uphole velocity
derive the static corrections depends on the four types of data survey. (b) Results of the survey. These surveys are useful
commonly available: the seismic data themselves, uphole for calibrating the results of refraction interpretations,
times from deep dynamite shooting, separate low-velocity particularly in areas where the low-velocity layer contains
layer (LVL) surveys, and uphole velocity surveys. one or more high-velocity layers.
Conventional seismic data are used in two very different
ways. There are refraction-based techniques which use the
first break information in a deterministic way to estimate the wishes to use a deterministic approach, to employ either an
near-surface model from which the static corrections are LVL or uphole survey to obtain near-surface velocity infor-
computed. Then there are the reflection-based techniques mation. Some acquisition geometries, such as crooked line,
which statistically derive residual static corrections to en- result in nonlinear first break data so that conventional refrac-
hance the coherency of the reflection. The long group lengths tion analysis of first breaks cannot be undertaken and it is then
and the source to near-receiver offsets of conventional seis- common practice to employ a separate LVL crew.
mic field geometries are usually such that the data fail to Uphole times from deep-hole dynamite shooting are an
adequately sample the uppermost layer(s), the zone in which additional source of information which is available at no extra
most variations of velocity and thickness occur, for determin- cost. This measurement provides an average vertical velocity
istic techniques (Figure 7). It is common practice then, when over the uppermost few tens of feet with one sample every
a deterministic approach is to be taken, to supplement the few hundred feet along the line depending on shothole depth
conventional seismic data with other information to help and spacing. Examination of adjacent uphole times from a
determine the near-surface model, or to combine refraction survey will give an indication of the consistency or variability
and residual techniques. when surface sources such as of the near surface. Unfortunately, these measurements,
vibroseis are used, the first breaks on the field records are which all too often’contain relatively large recording errors,
sometimes so poor (as in Figure 8) that it is imperative, if one do not allow us to define a detailed near-surface model and

THE LEADING EDGE JANUARY 1993 4;

 Figure 11. The floating datum is usually a running aver-
age of the topography over the length of the recording
cable.
Figure 12. Field static computations. Uphole time (tub) is
measured (except for surface source techniques where it
then the computed average velocities cannot be reconciled is zero). Shot elevation static Shot
with the near-surface velocities determined by other methods. weathering static Receiver
Separate LVL surveys enjoy enormous popularity in some elevation static Shot weathering
areas of the world (see Determination of static corrections by static The shot static, +
A.W. Rogers, in Developments in Geophysical Exploration , is subtracted from the seismic trace; the receiver
Methods, Applied Science Publishers, 1981). They are small static, +t is also subtracted. Weathering statics are
scale refraction surveys employing single geophones rather not always included and elevation statics are sometimes
than groups, with spread lengths on the order of a few tens of applied with only an approximate near-surface velocity.
feet to a few hundred feet. Figure 9 shows an example of a In such cases, any deficiency in the statics correction is
field geometry which would be repeated perhaps every one- remedied by the application of residual statics correc-
quarter mile. There is a range of different sources available t ions.
which are suitable, not only for LVL work, but also for
high-resolution seismic profiling. R.D. Miller et al. compared
them in Field comparison of shallow seismic sources (GEO-
PHYSICS 1986). It is necessary to ensure that the source is
powerful enough to provide adequate penetration and that a
long enough spread length is used such that the deepest zone
of investigation overlaps with the shallowest refractors on the
conventional data. Less attention needs to be given to the
choice of geophone since the response characteristics of the
geophones used in conventional reflection surveys (10, 14,
20, and 28 Hz) do not become nonlinear until around 500 Hz.
The data provided by these LVL surveys, then, fill in the zone
inadequately sampled by the first break information on the
conventional seismic record. Figure 13. Relationship between the surface topography,
In uphole velocity surveys, a small charge (a blaster cap the fixed datum, and the base of the drift layer (which may
or up to 4 oz of dynamite) is detonated at incremental depths be stratified). Drift statics compensate for all of the low
of say 20 ft in a hole from a total depth of say 300 ft up to the velocity layer, even that below the fixed datum.
surface (Figure 10). A geophone at the surface measures a
series of uphole times from which interval velocities are
calculated. By itself, an uphole survey doesn’t determine which contribute to the gather. The extension of this concept
layer boundaries with precision, so it is best used in conjunc- to every gather in the seismic line results in a variable or
tion with other information to ensure that measurements are floating datum which is a running average of the receiver
made close to horizon interfaces. The cost of uphole surveys elevations, the averaging taking place over a spread length.
usually limits the number conducted in a given survey area, With rugged topography and for 3-D surveys, it may be
and the resultant sparse spatial sampling makes them unsuit- desirable to generate a floating datum surface to avoid differ-
able for mapping lateral velocity variations and rapid changes ent datum elevations at line intersections. Such a surface is
in thickness of the near-surface layers. formed by averaging the topography elevations lying within
a circle of inclusion whose diameter is equal to the spread
length.
T he datum. Before static corrections can be computed and Since the interpreter would like all the seismic data in a
applied, it is necessary to select a suitable reference datum. project area to be referenced to the same horizontal datum to
Seismic reflection theory is based on a horizontal datum, but facilitate mapping, the floating datum is corrected to the
the datum of field data is the topography. A floating datum is horizontal datum after the NMO correction has been applied,
usually used until after the normal moveout (NMO) correc- often after the data has been stacked. Figure 11 illustrates the
tion as a compromise between reality and theory. relationship between topography, floating datum, and hori-
The floating datum is chosen so that the field statics are zontal mapping datum. Only in areas of very low topographic
kept small. This can be achieved by selecting a datum for each relief lying close to the horizontal datum is the floating datum
gather which is the mean of the elevations of the receivers dispensed with and the horizontal datum used in processing.

48 THE LEADING EDGE JANUARY 1993

F ield statics. As we have seen, from a locity layer is the glacial drift, or boul- tions in areas of glacial drift; it is often
complete knowledge of the locations der clay, which exhibits considerable applied to statics from any area where
and elevations of the shots and receiv- variation in velocity and thickness and there is a sharp irregular boundary be-
ers, and the velocities and thicknesses of overlies much higher velocity bedrock. tween low-velocity surface sediments
the near-surface layer(s), a complete Static corrections must compensate for and bedrock, particularly when the sur-
statics solution can in theory be ob- the whole of this low-velocity layer, face extends below the fixed datum.
tained. This computation is referred to even that below the fixed datum, to
as the field static to differentiate it from avoid serious distortions in the seismic
other static computations. In practice, it data. Parts II and III of this tutorial will be
is used to describe the initial static cor- Drift statics is now used more widely published in the February and March
rections made to the field data when than merely to describe the static correc- issues of TLE.
correcting the field records from the
topography to the floating datum on the
basis of either uphole velocity survey
results, LVL survey results, or the re-
fraction analysis of the first breaks. Some-
times it is no more than an elevation
static applied using some average ve-
locity which takes no account of varia-
tions in the near-surface velocity regime.
The uphole velocity survey as illus-
trated in Figure 10 yields a model of the
near-surface layers’ thicknesses and ve-
locities directly. The LVL survey re-
sults are analyzed by one of the refrac-
tion analysis techniques to yield the
model of the near-surface velocities and
bed thicknesses. Or the first breaks of
the seismic data themselves are ana-
lyzed directly to determine the near-sur-
face model.
Once the near-surface model has
been determined, then the elevation and
weathering corrections at both the shot
and receiver locations can be simply
determined as shown in Figure 12 where
the conventional situation of the datum
lying beneath the weathered layer is
shown. However, it is equally likely that
a floating datum, at least, will lie above
the topography or within the weathered
layer.
It is desirable to keep static correc-
tions as small as possible before the
NMO correction. In most surveys, this
is easy since the topography is fre-
quently of low relief, resulting in rela-
tively small field static corrections.
Where there is a thick low-velocity
layer or rugged topography or both, it
may be necessary to apply only an ele-
vation correction from the topography
to a floating datum, reserving a much
larger static until after NMO correction.

D rift statics. The term drift statics

arose to describe the static corrections
necessary to compensate for glacial
drift. In areas where glacial drift occurs
at the surface, the topography, drift, and
datums frequently exhibit the relation-
ships shown in Figure 13. In such areas
there is no dry aerated layer; the low-ve-
Part II

Static corrections—a review

Refraction statics. One of the reasons for deriving and collectively referred to as linear techniques, while those
applying static corrections is to ensure structural integrity in methods which allow rapid changes in refractor geometry and
the processed section. Static anomalies whose spatial wave- velocity are collectively referred to as nonlinear techniques.
lengths are longer than a spread length are not uncommon and In the real world, the topography is never flat, the refractor
if not corrected produce false structures in the seismic section is never planar, the velocity of the low-velocity layer proba-
(Figure 14). Refraction statics are effective for correcting bly changes both laterally and vertically due to compaction
these long spatial wavelength anomalies, and they are also and lithology variations, and finally the subweathering veloc-
effective against shorter spatial wavelength anomalies. ity probably varies laterally also. So, instead of the simple
Refraction methods allow us to derive estimates of the
thicknesses and velocities of the near-surface layers by ana-
lyzing the first breaks on the field records. There are many
methods which have been proposed over the years, but all are
based on the same basic principal of refracted energy.
The key concept in seismic refraction is that when a
seismic ray crosses a boundary between two formations of
different velocities, then the ray is bent according to Snell’s
law (Figure 15) which states that the ratio of the sines of the
incident and refracted angles is equal to the ratio of the
velocities of the two formations. As long as the velocity of
the lower layer is faster than the velocity of the upper layer,
the refracted ray will be bent toward the horizontal and there
will come a point, as the angle of incidence is increased, at
which the refracted angle is 90°. When this critical angle is
reached, the ray will travel horizontally in the second layer
close to the formation boundary. Energy traveling in the faster
medium close to the boundary continuously excites waves in
the upper layer which are transmitted back to the surface.
Energy refracted in the faster layer arrives at the surface
before the direct arrival, and it is known as the refracted or
headwave. This situation is shown in Figure 16 with the
time-distance plot of the first arrivals used to determine the
velocities of the layers involved (the inverse of the slope on
the graph). The thickness of the surface layer can be com-
puted from the slope and intercept-time values.
The most common extensions of the theory from the
situation depicted in Figure 16 are first to a multilayered
solution, then to a dipping layer solution, and finally to the
situation where the velocity of the top layer increases with
depth. (The reader who wishes to consider the formulas
associated with these and other extensions of the concept is
referred to An Introduction to Geophysical Prospecting by
MB. Dobrin or Seismic Refraction Prospecting, edited by
A.W. Musgrave, SEG, 1967.) Refraction techniques whose
theory is based on the assumption of planar refractors are

Part I of this three-part tutorial series dealt with the variabil-

ity of the near-surface velocity distribution and how it affects
the seismic data. Also considered were the types of data Figure 14. A seismic section showing the effects of differ-
available from which the near-surface velocity distribution ent statics correction technologies on the structural integ-
may be determined. The datum, field statics, and drift statics rity of the data. (a) A false time structure remains when
were discussed. Part II focuses on refraction statics. Next the residual statics technique fails to detect and compens-
month, Part III will cover residual statics, statics and velocity ate for a static anomaly of long spatial wavelength. (b) The
analysis, and some guidelines for the choice of method. anomaly is removed by a refraction technique.

FEBRUARY 1993 THE LEADING EDGE 115

 Figure 15. Refraction of a seismic ray Figure 16. Refraction of a seismic ray (a) at the critical angle of incidence and
at the boundary between two rocks the associated time-distance plot of first arrivals (b).
exhibiting different velocities.

relationship depicted in Figure 16, in which the velocities of

the weathered layer and subweathered layers are the inverse
of the slopes of a graph and the low-velocity layer thickness
is a linear function of the intercept time, we have the arrival
time of the first energy being some unspecified nonlinear
function of: thickness of the low-velocity layer, velocity of
the low-velocity layer, velocity of the subweathering layer,
and source/receiver separation. Those deviations of the sub-
surface from the theoretical linear model introduce errors into
the computation by linear techniques of both the thickness of
the weathered layer and the refractor velocity. In Inversion o f
refraction arrivals: A few problems (Geophysical Prospect-
ing 1990), L. Zanzi analyzed some of the errors and con-
cluded that by far the greatest errors came from curved
refractors at depth. He also concluded that, of the common
methods of refraction analysis, the technique developed by
D. Palmer in The Generalized Reciprocal Method of Seismic
Refraction Interpretation (SEG, 1980) appeared to be the
most accurate. It should be noted, however, that Zanzi did not
consider all methods, and the results that he presented were
based on a theoretical analysis of errors, illustrated with one
or two model examples. Figure 17. (a) A time-distance plot showing that refrac-
In general, the statics methods based on a theoretical linear tions along a dipping interface from opposite directions
model are derived from early techniques when calculations produce different slopes (giving apparent velocity) and
were done manually and, as a result, they derive static corre- different intercept times. (b) In LVL surveys, it is neces-
lations from the data without first deriving the near-surface sary to record the refractions shooting both updip and
low-velocity layer model. This may be achieved according downdip if velocities, dip, and depths are to be deter-
to: mined. The multiplicity of data in conventional multifold
seismic acquisition permits a plot such as (a) to be con-
T structed from data in the common receiver plane without
the need for reverse shooting in the field.
The terms are as defined in Figure 16. The techniques
based on this formula use different methods of estimating the
intercept time directly from the seismic data and rely on 1967). Figure 17 illustrates the extension of the method to a
the large level of redundancy in CMP data to overcome the single dipping refractor for which both forward and reverse
noise problems associated with automatic picking. profiles are illustrated. This basic method was originally
carried out by manual computation, along with extensions to
multiple layers and continuous increase of velocity with
S lope/intercept method. This technique adheres to the key depth within the different layers. The method now forms the
concepts of the seismic refraction method and is described in basis of a number of commercially available software pack-
some detail by W. A. Knox in Multilayer near-surface refrac- ages. These systems are limited to either a three- or four-layer
tion computations (Seismic Refraction Prospecting SEG, case and can handle dipping refractors.

116 THE LEADING EDGE FEBRUARY 1993

 Reciprocal methods. A number of reciprocal methods have
appeared over the years which have a common origin in the
method described by L.V. Hawkins in The reciprocal method
of routine shallow seismic refraction investigations (GEO-
PHYSICS 1961). The basis of the method is what Hawkins
called the time-depth term which is similar to the delay time
except that the surface and refractor are no longer assumed to
be approximately horizontal so that the depth terms of Figure
18 and are no longer vertical, but perpendicular to the
refractor. Differences in traveltimes over similar raypaths are
used to estimate the time-depth term and, hence, the intercept
Figure 18. The intercept time is the sum of two delay time and static.
times, one at the shot and the other at the receiver The Generalized Reciprocal Method of Palmer is probably
the most commonly used derivative method; it is illustrated
cos Delay time theory requires the surface and in Figure 21. Arrival times at two geophones, separated by
refractor to be approximately horizontal. what is termed the XY distance, are used in refractor velocity
analysis and time depth calculations. At the optimum XY
spacing (determined by tests), the forward and reverse rays
D elay time methods. Delay time is not an observable emerge from near the same point on the refractor, so the
quantity; it is defined as the time between the shot or receiver refractor need only be planar over a very small interval. The
and the refractor minus the time necessary to travel the normal depth calculation does not need an accurate determination of
projection of the raypath on the refractor. It is explained in the optimum XY spacing, but it is critical for the derivation
Figure 18. (For further details, see the articles by K.M. Barry of an average velocity and for the detection and accommoda-
and L.W. Gardner in Seismic Refraction Prospecting, SEG, tion of hidden (thin) layers and velocity inversions (a low-ve-
1967). The Fan technique (see A novel method of deriving locity layer beneath one of higher velocity).
weathering statics from first arrival refractions by G.W. Figure 22 illustrates the ABCD method described by M.S.
Hollingshead and R.R. Slater, paper presented at SEG’s 49th Bahorich et al. in Static corrections on the southeastern
Annual Meeting, 1979) and the Chronos method (see First Piedmont of the United States (GEOPHYSICS 1982). It directly
arrival picking on common-offset trace collections for auto- estimates the relative receiver static between two adjacent
matic estimation of static corrections by F. Coppens, Geo- geophone positions and is used in conjunction with tech-
physical Prospecting 1985) are based on the delay time. They niques that directly derive the absolute source and receiver
produce, from picked first breaks, estimates of the intercept statics at infrequent intervals. For example, one might have
time along with weathering and subweathering velocities at uphole surveys every mile along a line to provide absolute
every receiver location (Farr) or shot location (Chronos) from estimates of the static corrections and use the ABCD method
which the statics are computed. In the Farr technique (Figure to give variations in statics along the line between upholes.
19) first breaks are picked on common shot records and Approximations made in the formulation result in small errors
estimated from common-receiver gathers. Figure 20 in the differential statics estimates of up to about 5 percent.
illustrates the Chronos method where picks are made on The technique relies on the redundancy of the CMP method
common-offset gathers. to statistically obtain robust estimates of the differential stat-
Both techniques were designed to compute static correc- ics between adjacent geophone groups which are then tied to
tions for a constant velocity weathering layer of slowly the sparse absolute measurements to provide continuously
changing thickness overlying a refractor of constant velocity. varying static corrections along the line. The authors claim
When these conditions do not obtain, then unacceptable that the method also works well on crooked line shooting
errors arise in the computed statics. where the source and receivers are not in a straight line.

Figure 19. Determination of intercept time by the Farr technique, (a) first breaks are shifted according to which
results in (b) first breaks being approximately horizontal. Deviations due to changing refractor velocity and poor field
geometry are detected from these displays. If V is not known, the intercept time is estimated as shown in (c). The static
correction is computed from the intercept time (from Hollingshead and Slater).

FEBRUARY 1993 THE LEADING EDGE 117

 Figure 21. In the Generalized Reciprocal Method, receiv-
ers x and y are selected from the field data so that re-
fracted waves from source positions A and B emit from
the same point on the refractor (from Palmer).
Figure 20. In the Chronos method, the delay time at
position + n can be computed from picked traveltimes
between + n, and i + m. If dip is zero at position i + n
and if shot depth is zero, is equal to geophone delay at
position i + n.
T

where Z is the elevation of the earth’s surface above the

datum plane (from Coppens).

Figure ABCD method with the source-receiver config-

uration used to obtain data for static correction. The
relative ABCD static correction between point B and
point C is the traveltime from point G to point C.
(from Bahorich).

Tomographic methods. Previous approaches to computing

statistics used the first break information to derive a model of
the near surface based on some theoretical assumptions such
as horizontal surfaces or constant velocities. However, when
such linear methods are applied at various discrete intervals
along a line, one quickly appreciates that horizons are not flat
and velocities are not constant. Another approach to the Figure 23. The generalized linear inversion method is a
computation of statics is to assume a model, compute what tomographic scheme which allows variations in the thick-
the first breaks would be by ray tracing through the model, ness of all layers except the first.
and then modify the model in order to minimize the differ-
ences between observed and modeled first breaks. Such is the
tomographic approach. Russell and by Olson. The model allows for both bed thick-
D. Hampson and B. Russell called their version of this ness and horizontal velocity variations. The interpreter/pro-
approach generalized linear inversion (First-break interpre- cessor must specify how many layers are in the near surface
tation using generalized linear inversion, Journal of the and estimate their thickness and velocity. The method is
CSEG 1984), while K.B. Olson described an identical method ideally suited to those areas where the near surface is re-
which he called inverse modeling (A stable and flexible stricted to two or three layers whose parameters vary over a
procedure for the inverse modeling of seismic first arrivals, predictable and limited range.
Geophysical Prospecting 1989). W.N. De Amorin et al. used The numerical equivalent technique uses the concept of
a one-layer model in a tomographic approach which they tomography to derive velocities to attribute to cells of width
called numerical equivalent (Computing field statics with the equal to the geophone group interval, bounded on the top by
help of seismic tomography, Geophysical Prospecting 1987). the topographic surface and on the bottom by a horizontal
Figure 23 shows the scheme used by both Hampson and refractor at the base of the LVL (Figure 24).

118 THE LEADING EDGE FEBRUARY 1993

Figure 24. Model used by the numerical equivalent tech-
nique. Cell velocities are iteratively adjusted to minimize
the square of the error between observed and computed
first break traveltimes. Field statics are then computed
from the model (from de Amorin et al.).

Figure 26. Time-term technique4 (a) Source and receiver

statics model. (b) Grid of skewed (similar to the shift in
the Farr technique) first arrival times showing projec-
tions for statistical averaging to obtain shot and receiver
statics (from Chun and Jacewitz).

mine shallow complex structure (SEG Expanded Abstracts,

1985), N.R. Hill described a method for imaging irregular
refracting interfaces by the downward continuation of re-
fracted arrivals where they constructively interface. Recently,
Aldridge and Oldenburg ( Refractor imaging using an auto-
mated wavefront reconstruction method, GEOPHYSICS 1992)
have described a computerized version of Hagedoom’s orig-
inal approach (Figure 25).

Figure 25. (a) Superposition of the two reconstructed T he time-term technique. In all of the previous techniques,
wavefront systems. Dashed line is the locus satisfying the the statics are derived according to some model of the sub-
refractor imaging condition. (b) Comparisons of true surface. The time-term method of J.H. Chun and C.A. Jace-
(solid) and imaged (dashed) refracting interfaces (from witz (in The weathering statics problem and first arrival time
Aldridge and Oldenburg). surfaces, Abstracts for SEG 's 51st Annual Meeting 1981)
does not require that a model be generated. It derives the
statics from a statistical analysis of the first breaks.
The difference between the observed and computed first Referring to Figure 16, the total traveltime from shotj
break traveltimes is minimized by iteratively adjusting the to receiver can be thought of as
velocities. From this model, field statics are estimated. It
should be noted that the base of the LVL is held fixed, and
the velocities do not vary vertically, nor are there multiple where is the shot static, is the receiver static, is the
layers-hence, the numerical equivalence. The authors claim shot-receiver separation, and is the refractor velocity.
that the technique, which requires no supplementary field Figure 26 illustrates how the time-term technique concep-
acquisition procedures, produces acceptable static correc- tually views the data. The problem is set up as a system of
tions with little interpretational effort. linear equations and solved by Gauss-Seidel iteration to pro-
vide a surface-consistent solution. The redundancy of CMP
data leads to a number of estimates of and which
W avefront reconstruction methods. Another approach to statistically reduce any errors from the automatic picking of
determining the near-surface model is by imaging the data. the first breaks.
Early wavefront reconstruction techniques determined the Chun and Jacewitz actually measured the of Figure 16
position of the refractor from the intersection points of and applied a compensation factor, which is susceptible to
wavefronts for forward and reverse shooting (see The plus- variations in to estimate The authors indicated how
minus method of interpreting seismic refraction sections by both vertical and lateral variations in could be accounted
J.G. Hagedoom, Geophysical Prospecting 1959). for. Displays from the method provided interpretive informa-
In Downward continuation of refracted arrivals to deter- tion to help the user recognize a number of situations which

FEBRUARY 1993 THE LEADING EDGE 119

 affect the linear technique such as from the original slope/intercept meth- and only about three hours, on the same
changes in refractor elevation, high-ve- od. Most methods, though, produce al- data set, using the tomographic soft-
locity inclusions occurring in the low- most identical statics solutions. The ware. So far, refraction analysis has
velocity layer, and a variable refractor advantages of the more recent methods taken place in the natural domain of the
velocity. lie in their speed of application. For recorded data, the time-distance do-
example, Amoco recently compared a main. Recently, though, Refraction sta-
Refraction statics summary. Refrac- slope/ intercept method with a tomo- tics in the wavenumber domain by L.
tion analysis of first breaks, for the der- graphic method. The statics produced Zanzi and A. Carlini (GEOPHYSICS 1991)
ivation of a near-surface velocity model were almost identical. However, it took demonstrated that if the computations
from which static corrections can be the geophysicist three days to obtain his are undertaken after Fourier transfor-
calculated, has progressed a long way statics using the slope/intercept method mation, then much time is saved without
any reduction of accuracy. This reduc-
tion in computing time enables advan-
tage to be taken of the redundancy
inherent in CMP data to obtain a robust
solution.
As a deterministic technique which
aims to derive the near-surface velocity
model, refraction statics suffers from an
inability to detect velocity inversions (a
low-velocity layer beneath a high-ve-
locity layer, such as seen in Figure 10 in
Part I) although some of the specific
methods provide diagnostic material
which implies the presence of an inver-
sion. Another drawback is their inability
to resolve thin beds (known as the hid-
den layer problem); and the interpreta-
tion of velocity increases with depth
within a layer can be problematic with
some implementations. The inability to
deal correctly with velocity inversions
can limit the technique’s usefulness in
areas with high-velocity rocks at or near
the surface, but uphole velocity surveys
can be the ideal way to supplement re-
fraction data in such circumstances.
Many refraction techniques require
time-distance measurements over the
same interval in both directions to
uniquely determine the velocity and dip
of the refractor. In the split-spread tech-
nique, this is clearly not a problem. If
acquisition is in the end-on roll-along
mode, then using the redundancy of data
in the CMP method together with reci-
procity, both forward and reverse pro-
files can be constructed from single-
ended profiles.
A common statics problem is misties
between survey lines. This has serious
consequences for the structural integrity
of any interpretation made from the data
because the interpreter cannot correct
the misties in the absolute certainty that
his corrections are right. When using
deterministic methods, and refraction
techniques in particular, it is necessary
to ensure spatial consistency of the
near-surface model throughout the sur-
vey area-before computing the static
corrections-in order to avoid introduc-
ing misties.

120 THE LEADING EDGE FEBRUARY 1993

 Part III

Static corrections-a review

R esidual statics. The application of elevation statics, or has to be selected so that the window (1) covers a time zone
field statics, or field statics followed by refraction statics where primary events are dominant, (2) is long enough to
never leaves the seismic data completely free of static anom- cover a number of primary events, and (3) is reasonably deep.
alies. These “residual” static anomalies are due to unac- The calculated timeshifts are related according to
counted for variations in the low velocity layer. No matter
how well the deterministic technique may derive velocities
and thicknesses of the near surface, it leaves something to be where R is receiver static, i is receiver index, S is source static,
desired. There are two reasons for this: is source index, is an arbitrary time shift for CDP
• The model is a simplification of the geology resulting in
a tradeoff between thicknesses and velocities which result in
inexact static corrections.
• The static correction is an approximation for a more
complex problem.
Residual static anomalies are compensated for by using
statistical correlation techniques which seek to enhance the
quality of the stacked traces by first correctly aligning the
reflections.
Most residual statics techniques are surface consistent and
are based on the concept that the times on each trace consist
of a shot static, a receiver static, NMO, and residual NMO.
In order to separate the various effects, the stacking diagram
shown in Figure 27 is used to help classify the traces. All
traces from the “common receiver plane” contain the same
receiver static, all traces from the “common source plane”
contain the same source static, etc. Different surface-consis-
tent techniques vary in their approach to estimating shot and
receiver statics.
Figure 27. The four principal trace planes of the stacking
diagram (from Taner et al. 1974).
Linear traveltime inversion. The first step in this method
is to apply an approximate NM0 correction so that the
reflection events in each gather can be considered to be
misaligned due to a source static, a receiver static, and resid-
ual moveout. Next is to obtain a time shift for all traces in
each CMP gather in order to optimize the stacked trace.
Time shifts are calculated by a cross-correlation of traces
(Figure 28) either by computing all pair-wise cross-correla-
tions and picking a consistent set of delays or shifting one
trace at a time until the sum of the weighted semblances or
cross correlations is maximized. The window for correlation

This is the third and final part in a tutorial series on statics.

Part I examined the variability of the near-surface velocity
distribution and how it affects the seismic data. The types of
data available, from which the near-surface velocity distri- Figure 28. A and B are parts of two seismic traces showing
bution may be determined, were considered as were datums, a static shift between them. C is the autocorrelation of A
field statics and drift statics. Part II considered refraction and D is the cross-correlation of A with B. The time
statics. Part III deals with residual statics and with statics difference between the maxima on C and D is the mea-
and velocity analysis. Some guidelines to the choice of sured time shift used to estimate the source and receiver
method are also provided. statics.

210 THE LEADING EDGE MARCH 1993

gather, is the residual moveout at the CDP gather, and maximization, J. Ronen and J.F. Claerbout (GEOPHYSICS
is the source-to-receiver distance. 1985), rather than relying on the picking of a cross-correlation
This can be visualized in an analogous manner to Figure maxima as the most likely criterion for determining the static,
26 (in Part II) for the time-term technique. The above expres- hypothesizes that static shifts should be determined so that
sion gives rise to sets of simultaneous equations which have the sum of the squares of the final stack is maximized. The
to be solved to obtain the statics for application to the data. procedure is shown diagrammatically in Figure 29 and is
However, it so happens that the sets of equations are over- explained as follows:
determined (i.e., there are more equations than unknowns) A supertrace is built from all the traces of the shot profile
and underconstrained (i.e., there are more unknowns than in sequence (trace F in Figure 29) which is then cross-corre-
there are independent equations). In Estimation and correc- lated with another supertrace of all the traces in the relevant
tion of time anomalies (GEOPHYSICS 1974), part of the stack, in sequence, without the contribution of that
M.T. Taner et al. assume that the arbitrary time shifts are shot (trace G in Figure 29).
zero (thus ensuring that the datum is not moved) in order to The maximum of that cross-correlation is then picked and
determine values for and S. G. Saghy and A. Zelei, in the stack corrected, the process being repeated for each shot.
Advanced method for self-adaptive estimation of residual A supertrace is then similarly built from all traces in the
static corrections (Geophysical Prospecting 1975), assume receiver profile, in sequence, which is then cross-correlated
0 and 0 but introduce limit values and weighting with another supertrace of all the traces in the relevant part of
factors computed from the data itself. They introduce and the stack, in sequence, without the contribution of that re-
employ an iterative process for estimating the time shifts. ceiver.
Wiggins et al., in their 1976 paper, use an iterative Gauss-Sei- The maximum of that cross-correlation is then picked and
del approach to obtain a solution which results in 0, thus the stack corrected, the process being repeated for every
moving the datum in an unconstrained way. receiver.
An optional constraining routine may be included in the
procedure to remove linear trends and glitches from the
Stack power maximization. In the linear traveltime inver- estimated statics.
sion method, cross-correlation is a nonlinear operation sus-
ceptible to failure in the presence of ambiguities or noise. The
ambiguities may be amplitude or phase distortions of the data Nonlinear inversion. In the linear traveltime inversion
due to intragroup static anomalies or variable source coupling techniques, the residual static is estimated by linear inversion
for which Taner et al. offer a solution in Static corrections.= of observed traveltime deviations, a process which involves
Time, amplitude, and phase (SEG Expanded Abstracts, minimizing an error function. A crucial component of the
1991). The stack power maximization method described in approach is to pick the delay times accurately. Gross errors
Surface-consistent residual static estimation by stack-power in picking the traveltimes are known as “cycle skipping” or

Figure 29. Supertrace cross-correlation as used in the Figure 30. In the simulated annealing technique, the stack
stack-power maximization technique. The plane contain- power, over a selected window of the trace, is normalized
ing the shot profile in the unstacked data volume is mov- with respect to the input stack and plotted as a function
ing up or down according to the static shift of the shot. of iteration. Static corrections are selected from probabil-
The CMP stack is changing as a function of that static ity distributions for the data and applied according to
shift. Maximizing the power of the CMP stack as a func- certain rules. It is seen that the power initially decreases
tion of that particular shot static is equivalent to maximiz- very quickly, but convergence towards the solution starts
ing the cross-correlation between two supertraces built to occur, in this example, at about the 1000th iteration.
from the shot profile (trace F) and the relevant part of the Each step in the function is equivalent to the method
CMP stack (trace G). The procedure is repeated for every finding a local maximum while it continues to search the
shot and geophone. Convergence is usually achieved model space (of static corrections) for the global maxi-
within 5-20 iterations (from Ronen and Claerbout). mum (from Rothman).

MARCH 1993 THE LEADING EDGE 211

 “leg jumping” and occur through a lack of adequate data are then correlated with the model trace. The traveltime
conditioning either when excessive noise is present or when difference between the model trace and all of the traces within
the choice of maximum on the cross-correlation function is the CMP is determined by the cross-correlation method.
obscure. This cycle skipping causes the minimization process These time differences, when applied to the traces within the
of the linear inversion to result in a local rather than the CMP, provide an optimum alignment of reflection events
desired global minimum. In Nonlinear inversion, statistical within the CMP for stacking purposes.
mechanics and residual statics estimation (GEOPHYSICS This technique should be used judiciously to remove any
1985), D.H. Rothman showed that the estimation of large remaining static anomalies of very short spatial wavelength
statics in noise contaminated data is better handled as a or to remove residual moveout from observed reflections for
nonlinear inversion problem, at which time cycle skips appear the purpose of enhancing the visual quality of the stack. It is
as secondary minima. relatively easy to abuse this technique, as illustrated in Figure
In Rothman’s technique (see Automatic estimation of 31 where a number of false reflections have been generated
large residual static corrrections, GEOPHYSICS 1986), ex- out of noise.
plicit cross-correlation of traces is determined and then, in-
stead of picking the peaks of these functions, the method
transforms them into probability distributions. A simulated Residual statics summary. The main advantages of resid-
annealing algorithm is then used which is in turn based on a ual techniques are that they are automated and that supple-
Monte Carlo technique in which random numbers are drawn mentary information such as that provided by LVL surveys
from the created probability distribution and used to itera- is not required. However, there are drawbacks due to the
tively update estimates of the statics until convergence to the nonuniqueness of the solution. The most serious of the disad-
optimal stack is achieved. The stack power described by vantages is that these techniques are not capable of correcting
Ronen and Claerbout (GEOPHYSICS 1985) is used as the long spatial wavelength static anomalies (anomalies of wave-
criterion for determining the quality of the solution and a plot
of stack power against iteration is used as a method of
visualizing when the process has gone far enough. The. best
quality solution occurs when the stack power is maximized
(Figure 30).

S ubsurface consistent. If the residual statics anomalies are

of short enough spatial wavelength to be contained within a
gather, then subsurface consistent statics (also known as trim,
correlation, and CDP consistent) can be derived and applied.
The method makes no attempt to apportion a time shift
between source and receiver and needs to be used wisely.
Subsurface consistent statics is a CMP process in which a
model trace is built for every CMP, usually by summing
together 5 or 7 stacked traces. All of the traces within the CMP

Figure 32. A.H. Booker et al. showed (in Long wavelength

static estimation. GEOPHYSICS 1976) that the effects of long
period statics and structure can he distinguished on the
Figure 31. Surface-consistent statics applied (top). The basis of the stacking velocity in the residual statics tech-
same line (below) with nonsurface-consistent or subsur- nique. Comparison of the central and lower parts of the
face-consistent statics applied. Note flat-lying reflectors figure demonstrate why proper static corrections are im-
on the west end and strongly dipping reflectors on the portant to velocity analysis, particularly if the velocities
eastern half of the line where there were none before. are to he used for anything other than stacking the data.

212 THE LEADING EDGE MARCH 1993

length longer than a spread length) as illustrated in Figure 14 mis-ties into a data set, although this risk can be minimized
(in Part II). The presence of these static anomalies can be if the data are preconditioned by scaling, filtering, etc. and if
differentiated from genuine structures on the basis of the the residual static anomalies are small.
behavior of the stacking velocities which fluctuate wildly due Finally, subsurface consistent statics can be misused to
to the presence of the static anomaly but may not change create coherent events and manufacture structural anomalies
perceptibly in the presence of a structure as seen in Figure 32. within the time gates used for correlation. The smaller these
The nonuniqueness may also result in an effect known as time gates, the more likely it is that an event will be created
decoupling which produces a sawtooth effect in the data by aligning waveforms on the traces within a CDP. Multiple
(Figure 33) and is a failure of the method to detect and correct passes through the process, even with statics limited to ±4 ms
short wavelength anomalies (static anomalies whose spatial on each pass, will exacerbate the process.
wavelength is less than the shot point interval). In their 1976
article, Wiggins et al. explain the origin of the phenomenon
and show that decoupling occurs whenever every nth receiver S tatics and velocity analysis. The processes of static cor-
location is occupied by a source (n being greater than 1). If rection and velocity analysis are inextricably linked and what
shooting every group is impractical or too expensive, then by ought to be a single nonlinear process is treated as a set of
varying the sequence of shots, the system of simultaneous
equations which has to be solved may be coupled; for exam-
ple, if shooting every other group, it is only necessary to
ensure that two shots occur on adjacent groups somewhere
along the line. In 3-D acquisition, P.W. Johnson patented
(1987) a novel field geometry to ensure that the problem of
decoupling would not arise. When the problem is discovered
in previously acquired data, many approaches are possible but
probably the best way is to compute surface consistent re-
ceiver statics from the two or more decoupled systems of
equations and then a single surface consistent receiver static
profile is computed for the line as a running average of the
various decoupled solutions.
Another problem is that serious static errors can appear in
the data as faults and, in such cases, if the errors are greater
than half the length of the seismic wavelet, “cycle skipping”
or “leg jumping” occur; i.e., the correlation is made with the
wrong cycle when residual static corrections are applied. This
results in the data containing traces that are misaligned by a
whole cycle of the seismic wavelet. This can be dealt with in
a number of ways: (1) revise the earlier static solution, (2) low
band-pass filter the data prior to residual statics, and (3) use
one of the techniques not susceptible to the problem.
Yet another problem is that there is no way of ensuring
with certainty that the residual technique will not introduce

Figure 34. (a) Schematic gather before NM0 showing

hyperbolic moveout curves with respect to the surface and
the datum. (b) The same gather after NM0 correction
showing the residual moveout which occurs when a large
Figure 33. Effect of decoupling on data quality. static shift is applied before the correction.

MARCH 1993 THE LEADING EDGE 213

 linear processes in an iterative manner. which keeps the larger part of the elevation static separate for
The concept of stacking velocity analysis (the most com- application after stack.
mon form of velocity analysis applied to seismic data) is In the same way that the elevation static is split into small
based upon a horizontally layered earth where the NM0 can and large components, so shot and receiver weathering statics
be approximated by a hyperbola. Any deviations of the to- are often separated into two components for the same rea-
pography or base of the LVL from horizontal and any varia- son-the effect on the velocity analysis of shifting the origin.
tions in weathering and subweathering velocities will cause The large component of the weathering static is twice the
the NM0 to deviate from near hyperbolic. Erratic deviations mean of the shot and receiver statics at a particular location,
in the moveout, due to variations of short spatial wavelength, and this is applied after the velocity analysis, often after stack,
cause the criteria by which the correctness of the velocities is while deviations from the mean of the shot and receiver statics
judged to perform badly, thus making velocity picking (which at a particular surface location are applied before velocity
is still an interpretive process) difficult. Longer wavelength analysis.
static anomalies can have dramatic effects on the values of As a result of applying the larger part of the static correc-
the velocities, as illustrated earlier in Figure 32. It is for these tions after stack, the stacking velocities and seismic data have
reasons that static corrections are applied before NM0 veloc- different reference surfaces and this difference, illustrated in
ity analyses. It should be noted that static anomalies whose Figure 35, must be taken into account if the stacking veloci-
spatial wavelength is greater than a spread length are far less ties are to be used for any other purpose, particularly depth
disruptive in terms of velocity analysis. Only if the anomaly conversion. The situation is exacerbated if the statics take into
ends abruptly do velocities behave as in Figure 32. account low velocity section from below the reference datum.
The datum of our seismic field data is the topographic Every time a set of static corrections is computed and
surface, and it is to this surface that the NM0 correction and applied to the data, the origin of the data (i.e., zero time) is
stacking velocity must be referenced. It can be demonstrated moved for velocity analysis purposes and so for those tech-
that application of a large static correction before the NM0 niques where analysis takes place after the application of an
correction results in a moveout curve which deviates from the NM0 correction, it becomes necessary to reanalyze the ve-
desired hyperbolic moveout curve. As shown in Figure 34, locities with every iteration. This reanalysis may well happen
this deviation of the moveout curve is not the erratic deviation anyway since, in processing, the efficacy of a process or
due to topography or the LVL but a systematic deviation due parameter change is judged by the effect it has on the quality
to a shift in the origin of the hyperbola’s coordinate system. of the stacked (or final) section. Thus, we see that the whole
The curve is no longer hyperbolic with respect to the trans- iteration process of static corrections and velocity analyses
lated origin, even though it is still a smooth curve, and fitting becomes an interpretational process in which the geophysicist
hyperbolic moveout results in the familiar “gull’s wing” must be involved since optimization of the data quality de-
residual moveout. The stacking velocities obtained by con- pends on its ultimate purpose.
ventional velocity analysis deviate greatly from the root mean
square velocities of the earth model which they are assumed
to approximate, particularly in the shallow section, although C hoice of techniques. While the concept of static correc-
the discrepancy between the two sets of values decreases with tions is simple enough, there are many interrelated complex-
depth. In order to keep the residual moveout as small as ities to be taken into account. In addition, there is a bewilder-
possible (for residual moveout acts as a high cut filter on the ing array of different methods which can be employed. So
data in the same way that statics do), it is necessary to keep where does one start? At the planning stage.
the static corrections as small as possible before the NM0 It is no use waiting until the data are acquired, the tapes
correction and stack. This is the main justification for pro- are at the processing center, and the crew demobilized before
cessing the data with the floating datum described previously, thinking about static corrections. Existing data may indicate
the types of static problems that could be encountered; topo-
graphic maps or aerial photographs may give a clue-rock
outcrops, alluvial valley, swamp, etc. If feasible, the area
should be scouted. The availability of, or choice of, crew may
determine the data types one can obtain and the statics tech-
niques to be used. Consideration also has to be given to the
software available for processing.
There are some general guidelines, however, to help the
final selection. We have seen that the better the field statics,
then the better conditioned the data will be for the next stage.
In some circumstances, the statics from an LVL survey fol-
Figure 35. When the larger part of the static corrections lowed by residual statics may well be acceptable. Or, no LVL
are applied after velocity analysis, the processed seismic crew may be available so one may have to settle for field
data will be referenced to the datum plane; the velocity statics being no more than elevation statics, leaving the ques-
functions will be referenced to the topography when the tion “what next?” wide open.
base of the LVL is above the datum plane, and to a Figure 36 (from F. Coppens’ 1985 paper cited in Part II)
separate reference surface (controlled by the statics) shows that a deterministic approach (refraction statics) fol-
when the base of the LVL is below the datum plane. These lowed by a statistical approach (residual statics) produced the
different reference surfaces have to be taken into account best result. This, however, does not mean that examples
when using seismically derived velocities for depth con- cannot be found where residual statics on their own produced
version, the best result. The general conclusion is not surprising since

214 THE LEADING EDGE MARCH 1993

we saw that the strength of the refraction based methods was Beyond conventional statics. Conventional static correc-
an ability to correct static anomalies of spatial wavelength tions are part of a continuum of corrections, really dynamic
greater than a spread length (thus ensuring structural integ- in nature, stretching from spatial wavelengths of less than a
rity). We see in Figure 36 that after the refraction statics, the geophone group length to those whose wavelength is several
section is still noisy; this is typical and is due to residual spreadlengths and which range in magnitude from a fraction
anomalies left by imperfections in the model. of a millisecond upwards in magnitude. As we encounter
We also saw that the strength of the surface consistent static anomalies at the extremes of the continuums, new
residual techniques lay in statistically adjusting the shorter problems are encountered which conventional static correc-
wavelength anomalies which manifest themselves over a
tions do not or cannot handle well or because the static
number of gathers and that the subsurface consistent tech-
niques provide a cosmetic finish, taking care of the shortest anomalies manifest themselves in other ways.
wavelength anomalies remaining in the individual gathers Static anomalies whose spatial wavelengths are less than
after the other techniques had been applied. Figure 36 shows that of the geophone group length are generated in the top 10
how much improved the data are when refraction statics are ft or so of the earth’s surface. The resulting intragroup statics
followed by residual statics. This approach should ensure the cause variations in the amplitude and phase of the seismic
structural integrity of the data, enable the data set to be data from trace to trace in the field records. In his 1991 article,
processed with no internal misties, and optimize the image Taner suggests how variations in the amplitude and phase of
quality for stratigraphic information. the data can be treated in ways similar to the way he treats

Figure 36. (a) Seismic line with only elevation statics applied. (b) Same line with elevation and refraction statics, (c) with
elevation and residual statics, and (d) with elevation, refraction, and residual statics (from Coppens).

MARCH 1993 THE LEADING EDGE 215

 residual statics. Such an approach may within the low velocity layer (which is by intragroup statics alone. When such
not restore the attenuated high frequen- equivalent to a very small static time phase changes occur spasmodically in
cies (Figure 6) which at the moment are correction) gave rise to an apparent seismic data, they are often considered
best captured by single geophone phase change in the seismic data (which to be, and treated as, a residual statics
(rather than extended array) recording. they explained as being due to trapped problem and are compensated for by a
Such a solution poses new problems, reverberations). Whatever the cause, if time shift rather than phase rotation of
however. such phase changes occur on an in- the data.
The 1979 paper by Ziolkowski and tragroup basis, they will give rise to In the marine environment, rugged
Lawill, cited in Part I, showed that the even more variability in the phase and seafloor topography will induce effects
termination of a thin high velocity bed amplitude of the signal than that caused in the seismic data not unlike severe
statics. J.A. Berryhill, in Submarine
canyons: Velocity replacement by wave
equation datuming before stack, and
O.Yilmaz and D. Lucas, in Prestack
layer replacement (wave-equation da-
turning)--both in GEOPHYSICS 1986-
use prestack layer replacement and
wave equation datuming (downward
continuation of the wavefield) to correct
the data.
A similar approach was used by C.
Beasley and W. Lynn in The zero-veloc-
ity layer: Migration from irregular sur-
faces (SEG Expanded Abstracts, 1989)
to do prestack depth migration from a
datum above all topography and infill-
ing with a zero velocity layer, to image
data from extremely rugged topography
where conventional statics and process-
ing fail. Such an approach, however,
needs a more detailed velocity field than
is usually available in areas of complex
shwture which are usually associated
with rugged topography.

Acknowledgments: The author would

like to thank Brian Russell for his in-
sightful comments on an internal
Amoco document which led to the writ-
ing of this series, Jin Lin (Amoco Pro-
duction Company) for the example of
subsurface consistent statics (Figure
31), and many other Amoco geophysi-
cists on whose work he has freely drawn
for other material.

Dave Marsden is
a consulting geo-
physicist with
Amoco (UK) Ex-
plorotion Com-
pony. He obtained
a PhD ingeophys-
ics from the Uni-
versity of Leeds
(1972) and joined
Amoco in 1976. At Amoco he has com-
pleted various assignments as inter-
preter, supervisor, researcher and con-
sultant, working primarily in Europe,
but also in North America.

216 THE LEADING EDGE MARCH 1993