Raef 2009

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Journal of Earth Science, Vol. 20, No. 3, p.

640–648, June 2009 ISSN 1674-487X


Printed in China
DOI: 10.1007/s12583-009-0053-9

Land 3D-Seismic Data: Preprocessing Quality


Control Utilizing Survey Design Specifications,
Noise Properties, Normal Moveout,
First Breaks, and Offset

Abdelmoneam Raef*
Geology Department, Kansas State University, Manhattan, Kansas, USA; Formerly, Kansas Geological Survey,
The University of Kansas, Lawrence, USA

ABSTRACT: The recent proliferation of the 3D reflection seismic method into the near-surface area of
geophysical applications, especially in response to the emergence of the need to comprehensively char-
acterize and monitor near-surface carbon dioxide sequestration in shallow saline aquifers around the
world, justifies the emphasis on cost-effective and robust quality control and assurance (QC/QA) work-
flow of 3D seismic data preprocessing that is suitable for near-surface applications. The main purpose
of our seismic data preprocessing QC is to enable the use of appropriate header information, data that
are free of noise-dominated traces, and/or flawed vertical stacking in subsequent processing steps. In
this article, I provide an account of utilizing survey design specifications, noise properties, first breaks,
and normal moveout for rapid and thorough graphical QC/QA diagnostics, which are easy to apply
and efficient in the diagnosis of inconsistencies. A correlated vibroseis time-lapse 3D-seismic data set
from a CO2-flood monitoring survey is used for demonstrating QC diagnostics. An important
by-product of the QC workflow is establishing the number of layers for a refraction statics model in a
data-driven graphical manner that capitalizes on the spatial coverage of the 3D seismic data.
KEY WORDS: preprocessing, quality control, 3D seismic, 4D seismic, trace header, geometry, vertical
stacking.

INTRODUCTION storage capacity (Damen et al., 2006; Li et al., 2006;


The application of 3D seismic reflection to near- Arts and Winthaegen, 2005). The requirements and
surface site characterization and monitoring injectates means of seismic quality control and quality assurance
has gained momentum in the light of increased interest (QC/QA) have been addressed by several authors (Li
in the established viability of geological sequestration et al., 2008; Bachrach and Mukerji, 2004; Jin et al.,
of carbon dioxide and subsurface hydrocarbon gas 2004; Thomas et al., 2004; Widmaier et al., 2002),
mostly for either the acquisition stage or for main
This study was supported by the U.S. Department of Energy processing and post-processing volumes. Many types
(No. DE-FC26-03NT15414). of processing problems arise from defected preproc-
*Corresponding author: abraef@ksu.edu essing steps such as incorrectly setting up field ge-
ometry (Yilmaz, 1987). A complete preprocessing QC
Manuscript received December 15, 2008. should include geometry check, equipment test, seis-
Manuscript accepted February 19, 2009. mic data quality, and seismic signal quantification
Land 3D-Seismic Data 641

(Burger et al., 1998). Seismic data quality in preproc- processing and overall cost when identified later in the
essing has been focused upon by Burger et al. (1998). processing flow. In this work the aim is to highlight
In this article, graphical QC techniques for the 3D preprocessing QC techniques that were used in the
seismic data preprocessing stage are presented; special course of time-lapse seismic monitoring of an en-
emphasis is directed to detecting errors in field ge- hanced oil recovery CO2 injection site in Central Kan-
ometry assignment. Since preprocessed seismic data sas, USA. The highly scrutinized preprocessing data
visualization in one data attribute domain (e.g., field have been gathered with the aim of imaging subtle
file Id-Channel No.) may not manifest any inconsis- reservoir seismic anomalies, hence the significance of
tency or errors. In the approach of this article the de- tight standards in the QC approach. The following QC
sign of multi-domain field data visualization has been techniques should be used with the understanding that
a key element in detecting preprocessing data prob- they are interdependent and/or complementary, thus
lems. Technical and operational considerations such corroborating each other’s QC capacity. This interde-
as cost-effectiveness, abundance of good quality gath- pendency of many of the QC diagnostics, in this arti-
ers, and pressure for shorter turnaround time may en- cle, strengthens the effectiveness of identifying prob-
courage a tendency towards underestimating preproc- lem stations or errors in geometry assignment or
essing errors assuming that subsequent advanced loading to trace headers. In the following sections of
processing will overcome those problems. Seismic this article, as an example, several preprocessing QC
data-processing geophysicists most often receive land techniques are explained utilizing various field data
seismic data in the form of preprocessed (coordinates specifications of only one swath 3D dataset.
loaded and data vertically stacked) 3D-seismic data
from acquisition contractors. TECHNIQUES
The sheer volume of field data involved in 3D Survey Specifications and 3D-Geometry Verifica-
seismic surveys, coupled with the multi-domain visu- tion
alization capabilities of modern workstations, call for Survey design specifications and actual field data
favoring graphical QC plots in the preprocessing stage. gathering may not be in agreement as a result of field
In this article, I present QC/QA techniques for the surface conditions and morphology, as well as human
preprocessing of land seismic data. These QC/QA di- errors during the survey acquisition and preprocessing
agnostic techniques are designed to graphically iden- practices. Utilizing survey design, actual spatial sur-
tify problems caused by erroneous vertical stacking, vey locations, and seismic bin fold, we both validate/
geometry assignment, and positioning errors and de- correct the 3D-geometry assignment and identify
viations of design locations. Evaluating QC data- problem stations that are to be isolated for accounting
integrity and 3D-geometry header information after for survey and/or preprocessing errors. The survey de-
preprocessing is a very important step before embark- sign for the 3D survey used here to illustrate the use of
ing upon subsequent data-processing operations. Er- various QC plots, consists of modified brick source-
rors that may occur in the preprocessing stage mostly station lines, 100 m apart, and five receiver lines
arise from misallocating or mislabeling of stations, spaced at 200 m (group interval 20 m) single-patch
incorrect vertical stacking of shot records at each sta- deployment (Fig. 1) covering a survey area of 3.6 km2
tion, swapped receiver cable connection (flipped and yielding a maximum of 24-fold common midpoint
channels), and 3D-geometry assignment inconsisten- (CMP) coverage.
cies or mismatching. Positioning errors, missing stations, and/or ge-
In time-lapse projects, where subtle reservoir ometry error diagnostics can be indicated by anoma-
seismic signatures are sought, the great significance of lous bin-fold distributions (Figs. 1a and 1c) that are
conservative QC is essential for the entire data proc- not consistent with survey design. The application of
essing effort, particularly at the early stage of pre- binning grid rotation could conceal error-diagnosing
processing and geometry assignments. This is where CMP-fold anomalies (Figs. 1b and 1c). Thus, it is
errors are made, that adversely affect both subsequent recommended that the QC binning grid have no rota-
642 Abdelmoneam Raef

tion (inlines orthogonal to land survey azimuth/source decreasing and orientation that help confirm the com-
lines). Spatial map displays (Fig. 2) of in-line and pliance of grid-layout to the land seismic convention
X-line of the binning grid ensure/verify the correct- of in-lines parallel to or at acute angle (after rotation)
ness of grid orientation: the direction of increasing/ to the receiver lines direction.

Figure 1. Pilot 3D-seismic survey layout with north-south source lines spaced by 100 m and west-east re-
ceiver lines (group interval of 20 m) at spacing of 200 m; anomalous lower fold of coverage (a) indicated
sources that have been either dropped or mispositioned during geometry assignments; corrected geometry
(b) results in fold of coverage that is consistent with survey design; rotating binning (c) grid can make it
difficult to see fold inconsistencies.

Figure 2. Cross-lines (left) and in-lines (right) spatial map displays resulting from rotated binning.
A matrix plot of “offset map” (Fig. 3) in the plane of source-channel in case of regular survey de-
Land 3D-Seismic Data 643

sign is an effective tool for pinpointing mislocation (blue=low, red=high) in each pin summarizes the
when analyzed in the survey-design source-receiver offset/azimuth distribution resulting from one shot line
relational framework. Moderate to high regularity of and one receiver line; observing patterns of change
survey design increases the potential of offset and can be a strong diagnostic tool for identifying missing,
azimuth matrix plots for QC purposes. The color scale mispositioning, and/or geometry assignment.

Figure 3. Change of offset pattern within a single patch (a) as a function of source-receiver layout (five re-
ceiver lines and nineteen source lines) with no irregularities; (b) irregularity/abrupt change in pattern is
indicative of geometry assignment errors (Channels 80–90, 110–120 and 16–170); (c) color scale manipula-
tions may enhance errors detection.

A matrix color display of source-receiver azi- facts that can be associated with azimuth variability.
muth (Fig. 4) can also be used for preprocessing QC In the case of a color-coded matrix plot of
purposes, where cross-examining the variations in source-receiver offset or azimuth displays, simultane-
offset and azimuth against survey specifications is in- ous examination is recommended; small channel-
formative in its consistency from one patch to the next mispositionings, particularly those parallel to receiver
or regularity from one receiver line to the next. Inter- lines extension, are likely to go undetected on offset
active workstation variations of offset/azimuth color maps whereas they are detectable on azimuth plots.
scale structure are very helpful for examining consis- Channel and source misallocation is distinguishable
tency and regularity issues that are basically to be ref- on offset and azimuth displays from the trend of any
erenced to, and compared against 3D-survey specifi- anomalous pattern that is perpendicular to the respec-
cations. A matrix color display of source-receiver tive axis. Inconsistencies identified would have to be
azimuth could also be helpful in the later stages of the resolved by either correcting mispositioning and/or
processing flow as it might help in recognizing arti- geometry assignment errors.
644 Abdelmoneam Raef

Figure 4. Positioning/geometry error-free pattern (a) of change of source-receiver azimuth within a single
patch (five receiver lines and nineteen source lines) as a function of source-receiver layout; irregular-
ity/abrupt variation (b) in pattern is indicative of geometry assignment errors; in this example pattern
change.

Noise
In general, source-generated as well as stationary
noise (Fig. 5), though harmful, the characteristic
properties of noise, such as moveout, location on re-
cord, frequency content and waveform, would allow
one to utilize those noise events in identifying cor-
rupted records and/or bad geometry assignments. This
is particularly true when examining data from the
same swath location in different time-lapse surveys,
where stationary noise shift (Figs. 5a and 5b) is in-
dicative of channel misnumbering. Misnumbering
would be undetectable from header plots or field file
ID (FFID) or trace numbering plots. This example
suggests that bad traces can be well utilized for posi-
tional and geometry QC purposes. A doubling of
source-generated airwaves and/or widening of sta-
tionary noise can indicate that two different records
have erroneously been vertically stacked and assigned
to the same station (Fig. 5c). Other multi-channel
source-generated noise can be utilized in a similar
manner to identify problematic stations. Utilizing
noise properties in a preprocessing QC technique
would help flag problem stations that would be sub-
Figure 5. Stationary noise (red arrow) at the right
jected to further corrective measures such as rectifying
position (a), shifted (b) due to mislabeling of chan-
channel numbering and geometry assignment, and re-
nels and changing in character (c) as a result of
applying vertical stacking of raw records.
erroneous vertical stacking which is also indicated
by duplication of source generated air wave noise.
Trace-Header Parameters and Normal Moveout
Trace header parameters and moveout (Fig. 6)
can be used effectively as preprocessing QC tools. In
Land 3D-Seismic Data 645

this study, we use surface receiver location number cording and source location number SRLN-misla-
(SRF SLOC-blue) and field file identification beling) is indicated by anomalous FFID oscillation
(FFID-red) plots against channel numbers besides (Fig. 6c). Traces with varying FFIDs (Fig. 6d) allo-
moveout for identifying problem records. Receiver cated to the same station, and anomalous airwaves,
line reversal (Fig. 6a; flipped cables) is indicated by indicate errors in vertical stacking. Data corruption is
both anomalous moveout and SRF SLOC reversal. indicated by both anomalous moveout and a plot of
Mislabeling of channels (Fig. 6b) is indicated by re- trace header parameters (Fig. 6e).
versal of moveout. Corrupted data (duplicate re-

Figure 6. Receiver line reversal (a) is indicated by


both anomalous moveout and source location num-
ber (SRLN)-trend reversal; mislabeling of channels
(b) is indicated by reversal of moveout; anomalous
SRFLN oscillation resulting from duplicate re-
cording and SRLN-mislabeling (c). Traces, with
varying FFIDs (d), allocated to the same station and
anomalous air waves indicating errors in vertical
stacking; anomalous moveout and trace header pa-
rameters signal data corruption (e).

First-Breaks resulting from very low signal-to-noise ratio. Thus,


Preprocessing QC application of first-break picks examining matrix displays of channel-source-first-
can benefit from advances in the speed and precision break pick times aids both fast editing of first-break
of the automatic picking capacity of modern process- pick times and identification of erroneous geometry.
ing platforms. Matrix displays of channel-source- A display of channel, offset, and offset-to-
first-break pick times (Fig. 7a), when compared with first-break pick time ratio (Fig. 7b) shows near con-
survey specifications (spatial relations of source-line stant values of offset-to-first-break times over ranges
receiver line), can also be utilized in a similar manner of offsets corresponding to refraction energy arriving
and compared with channel-source-offset plots (Fig. from two refractors. Such plots are informative in
3). Anomalous first-break pick patterns (Fig. 7a) re- terms of identifying errors in pick times and/or ge-
flect either geometry assignment errors or bad picks ometry assignments, and as a by-product establishing
646 Abdelmoneam Raef

Figure 7. Anomalous first-break pick ((a) dotted outlines) pattern reflects either geometry assignment er-
rors or bad picks that result from very low signal to noise ratio; display of offset-to-first-break pick time
ratio (b) is nearly constant over ranges of offset corresponding to refraction energy arriving from “con-
stant” refractor velocity traces/channels and SINs corresponding to anomalous regions have to be checked
for preprocessing and/or first-break picking errors. In (b) one could establish that a three-layer model
should be the choice for near-surface refraction statics.

Figure 8. Poorly predicted Y- (a) and X-values (b) outlined by ellipses can be isolated for further analysis
for relating the low quality of prediction to either bad picks or erroneous spatial allocation.

the number of layers to be used in building subsequent of refraction offsets, a stable ratio of offset-to-
external model refraction statics. Near-surface veloc- first-break pick time is indicative of consistent ge-
ity variability sampled by direct wave arrivals is mani- ometry and good quality time picks. In Fig. 7b one
fested directly on the offset-to-first-break picks at could establish that a three-layer model should be the
small offsets (Fig. 7b). In simple terms, within a range choice for near-surface refraction statics as manifested
Land 3D-Seismic Data 647

by three zones of near uniform x/t for first-breaks of processing due to multiplicity of potential sources of
the entire dataset. errors, the potential for utilizing this arsenal of diag-
A more quantitative approach to using first- nostics lies in its redundancy and the diversity of
break picks for quality control on source and receiver visualization domains. Very often 3D seismic data
locations is to utilize this system of equations acquisition is contracted to an external contractor, who
( XS − XRi ) 2 + (YS − YRi ) 2 = [(TFB − τ SRi )(V1i )]2 would also carry out the preprocessing phase. Such
where XS and YS are source coordinates; XRi and YRi situation favors implementing QC measures before
are the i-th receiver coordinates; TFB is first-break pick embarking on further data processing. In Fig. 7b, off-
time; τSRi is the sum of the source and i-th receiver de- set over arrival (color scale) in the plane of offset-
lay times (possibly from uphole times or from regres- channel No. for the entire dataset enables judging of
sion of first breaks on offset); and V1i is the first re- the consistency of first-break picks and also estab-
fractor velocity. A similar system of equations holds lishes an objective data-driven approach for establish-
for common receiver gathers. ing the number of layers for near-surface refraction
Plotting predicted spatial coordinates against ac- statics. This is an example of the double benefits of
tual or survey design values (Fig. 8) would help iden- applying preprocessing QC to field 3D seismic data-
tify badly predicted values, which can then be isolated sets and the usefulness of data redundancy aspects.
for further analysis to determine whether the low
quality of prediction results from bad first-break picks CONCLUSIONS
or erroneous spatial allocation. Geometry assignments, and positional and label-
ing errors, when they are carried through the initial
DISCUSSION stage of processing (preprocessing), have seriously
For many near-surface geophysicists, the han- detrimental effects on subsequent processing, particu-
dling of large volumes of field seismic data could be larly normal moveout and velocity analysis. Therefore,
overwhelming when it comes to preprocessing the re- it is very important to pinpoint affected records and
ceived field data. In this account, several graphical account for those errors. The underlying thought of
techniques of preprocessing QC/QA could be very this approach is to use various trace-to-trace as well as
helpful and rapid means of addressing problems that shot-to-shot data and survey relational parameters in
could otherwise jeopardize subsequent processing the process so that no or a minimal number of bad re-
steps. This stage (data preprocessing) of data prepara- cords would go undetected.
tion has been underemphasized, as is clear from the We noticed that most of the preprocessing errors
lack of reported workflows in the literature, particu- that were not related to data quality could be identified
larly as one would survey near-surface publications. before proceeding in the processing flow. Proper
To my knowledge, there has been no published mate- communication and/or good quality metadata would
rial documenting various options for applying graphi- help answer questions raised during the preprocessing
cal QC techniques to the preprocessing of 3D seismic stage. Many articles have reported on quality control
data, although errors due to, for example, geometry of data processing with the assumption that correlated
assignments are hard to detect when tackling a larger preprocessing data and their spatial header attributes
number of source stations as is the case in 3D and 4D are intact. In this article, we did not intend to cover an
seismic applications. Geometry assignment, although exhaustive list of “QC tools” for the entire processing
seemingly a simple and straightforward task in 2D flow but to focus on data submitted in the form of
datasets, could be very challenging in case of the very correlated records with geometry assigned. No as-
often-encountered large volume data sets in 3D and sumption was made as to the correctness of either
4D seismic applications. Thus the significance of ap- geometry assignment or positioning parameters. Our
plying rapid graphical techniques utilizing plots in methodology advocates using both data-extracted and
various domains. spatial survey specifications to validate geometry as-
Despite the challenges of 3D seismic data pre- signment, positioning parameters, and labeling attrib-
648 Abdelmoneam Raef

utes. Most of the QC techniques presented in this arti- Burger, P., Garott, A. R., Granger, P., 1998. Improving Resolu-
cle are applicable to multi-patch/swath surveys. tion and Seismic Quality Assurance through Field Pre-
processing. The Leading Edge, 17(11): 1562–1569
ACKNOWLEDGMENTS Damen, K., Faaij, A., Turkenburg, W., 2006. Health, Safety
Support for this research was provided by the and Environmental Risks of Underground CO2 Storage—
U.S. Department of Energy, project #DE-FC26- Overview of Mechanisms and Current Knowledge. Cli-
03NT15414. I thank Rick Miller for helping to im- mate Change, 74: 289–318
prove this article, Kansas Geological Survey geo- Jin, L., Chen, X., Li, J., 2004. New Quality Control Method for
physical field crewmembers for acquiring the data Time-Lapse Seismic Processing. SEG, Expanded Ab-
presented, and George Clark and Mary Brohammer stracts, 23(1): 2315–2318
for styling the text. I also thank the Kansas Geological Li, J., Chen, X., Zhao, W., et al., 2008. Purposeless Repeated
Survey for supporting this study. My thanks also go to Acquisition Time-Lapse Seismic Data Processing. Petro-
two anonymous reviewers, especially reviewer who leum Science, 5(1): 31–36
supplied constructive comments that helped in im- Li, R., Urosevic, M., Dodds, K., 2006. Prediction of 4D Seis-
proving the manuscript. mic Responses for the Otway Basin CO2 Sequestration
Site. SEG, Expanded Abstracts, 25(1): 2181–2185
REFERENCES CITED Thomas, J. W., Hufford, J. M., Hoover, G. M., et al., 2004.
Arts, R., Winthaegen, P., 2005. Monitoring Options for CO2 Acquisition/Processing-Teepee Technology Applied to the
Storage: Carbon Dioxide Capture for Storage in Deep Design of P-Wave Acquisition, Part 1—Introduction and
Geologic Formations. Elsevier Science, Amsterdam. Tutorial. The Leading Edge, 23(1): 16–19
1001–1013 Widmaier, M., Moe, J., Brittan, J., et al., 2002. Case Studies in
Bachrach, R., Mukerji, T., 2004. Portable Dense Geophone Quality Control for 4D Seismic Processing. SEG, Ex-
Array for Shallow and Very Shallow 3D Seismic Reflec- panded Abstracts, 21(1): 2090–2093
tion Surveying: Part 1—Data Acquisition, Quality Control, Yilmaz, O., 1987. Seismic Data Processing. Society of Explo-
and Processing. Geophysics, 69(6): 1443–1455 ration Geophysicists, Tulsa

You might also like