BUSINESS BOTTOM LINE

Coordinated by Lawrence Gochioco

The impact of seismic amplitudes on prospect risk analysis

ROCKY RODEN, Rose and Associates, Centerville, Texas, USA
Downloaded 05/30/14 to 130.207.50.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

MIKE FORREST, Rose and Associates, Dallas, USA

ROGER HOLEYWELL, Marathon Oil Corporation, Houston, USA

E ssentially all companies involved in oil and gas explo-

ration and development must account for the various geo-
logic risk factors associated with their specific prospects.
Since seismic data (calibrated with well control if available)
are a primary interpretation tool to determine these risk fac-
tors, the presence of seismic amplitudes that are potentially
associated with oil or gas pays is extremely important.
However, interpreters evaluating prospects have had to
inherently know how seismic amplitudes impact the geo-
logic chance factors and ultimately the probability of drilling
success (Pg).
From 2001 to 2004, a DHI Risk Analysis Consortium of
oil companies has been working to systematically quantify
how seismic amplitude anomalies impact predrill estimates
of Pg. This goal was accomplished by addressing specific
seismic amplitude characteristics in different geologic envi-
ronments, and by quantifying the quality of the data from
which these interpretations are made. Figure 1. Interpretation workflow concept.

Background. The conventional approach to determine Pg,

which represents the chance of finding hydrocarbons in a
reservoir capable of sustained flow, is to evaluate the geo-
logic components required for a hydrocarbon reservoir to
exist. The number of chance factors and specific definitions
may vary slightly from company to company, but a com-
monly applied system of five factors is source, migration,
reservoir, closure, and containment (Table 1). However, the
presence of a seismic amplitude anomaly may have a sig-
nificant influence on the risking of the geologic chance fac-
tors.
For example, the presence of a potential hydrocarbon-
generated seismic amplitude anomaly may imply a work-
ing petroleum system, which would be strong evidence
supporting the presence of source. A clearly defined seis-
mic amplitude area also suggests that migration and trap-
ping of hydrocarbons has occurred, that trap seals may be
sufficient, and (depending on the geologic setting) reservoir
rock may be present.
The above logic is of course “circular” in that the pres-
ence of an amplitude anomaly does not necessarily indicate Figure 2. AVO classes based on amplitude change with offset from the top
it came from a hydrocarbon-charged reservoir. A better of gas sands (Rutherford and Williams, 1989; Castagna et al., 1998).
approach is to assess the confidence level that a seismic
amplitude anomaly is truly generated from the presence of The methodology developed by the DHI Consortium
hydrocarbons by asking the following questions: addresses all these issues with an interpretation workflow
concept (Figure 1) that has proven systematic, consistent,
• How many HCI or DHI characteristics does the anom- and produced very realistic results based on a calibration
aly display? data set of almost 100 drilled wells.
• Should an amplitude anomaly of this type even be pre-
sent in this geologic setting? Classifying seismic amplitudes. To determine which seis-
• In this geologic setting, should there always be an ampli- mic characteristics are pertinent for specific geologic trends,
tude anomaly associated with trapped hydrocarbons? it is necessary to classify hydrocarbon-generated amplitude
• Are the seismic data being used to evaluate the ampli- anomalies based on their geologic setting. The consortium
tude feature of sufficient quality to make reliable inter- chose the classification approach of Rutherford and Williams
pretations and valid conclusions? (1989) based on three classes of AVO responses from the top
of gas sands. With the addition of Castagna’s (1998) class 4,

706 THE LEADING EDGE JULY 2005

 essentially all geologic settings
are accounted for in clastic envi-
ronments (Figure 2).

• Class 1: Very consolidated

sands (porosity <15%) that
typically have gross interval
velocities >12 500 ft/s (3650
Downloaded 05/30/14 to 130.207.50.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

m/s). These reservoir sands

have higher impedance than
the encasing shale with the
density contrast having a
greater effect on the AVO re-
sponse. These reservoirs are
usually found onshore in
hard rock geologic settings
(e.g., mesozoic and paleozoic
rocks).
• Class 2: Less consolidated
sands than class 1, but more
consolidated than class 3.
These moderately compac-
ted sands (porosity 15-25%)
are found offshore and on-
shore. Gross interval velo-
cities of the sands are
typically between 8500 and Figure 3. Amplitude characteristics for three categories of class 3 anomalies.
12 500 ft/s (2650-3650 m/s).
The acoustic impedance of the gas sand and the encasing
shale are about equal.
• Class 3: This is the typical “bright spot” or “DHI” setting
where unconsolidated reservoir sands are encased in
higher impedance shales, usually of Tertiary age. These
reservoirs commonly have porosities >25% and contain
gas or high GOR oils. Gross interval velocity of these sands
is usually <8500 ft/s (2650 m/s). Class 3 targets were ini-
tially found exclusively offshore but, due to higher qual-
ity seismic data, are now being observed more frequently Figure 4. Locations of drilled prospects in the DHI Consortium database.
onshore.
• Class 4: Unconsolidated sand similar to class 3, but over-
lain by high velocity hard shale, siltstone, or carbonate.
Shear-wave velocity in this overlying hard interval is
larger than that of the gas sand (this situation is opposite
for all other classes).

It is important to realize that the boundaries between

classes are gradational and therefore it may be necessary to
evaluate amplitude anomalies in two classes.
To put the prospect in its proper geologic context and
to preserve the case for future comparisons, it is necessary
to document the prospect’s location, well class (e.g., devel-
opment, wildcat), type of closure, terrain (onshore/offshore),
depth-to-target, water depth, expected reservoir size and
thickness, etc. The expected reservoir geologic characteris-
tics such as age, lithology, lateral geometry, and vertical
boundaries are also input along with the presence and loca-
tion of analog fields and discoveries.

Assessing the base geologic risk. The interpreter estimates

initial Pg based on regional and local geologic interpreta-
tion which ignores the seismic amplitude anomaly. Thus,
the initial Pg is based on environment of deposition stud-
ies, reservoir isochore (isopach) maps, structural and/or
stratigraphic maps from wells and seismic, etc. In other
words, the geologic chance factors in Table 1 are risked with-
out the seismic amplitude anomaly being considered. pure stratigraphic trap where the location of the prospect
Typically, initial Pg will be very low if the prospect is a is only identified by the amplitude anomaly and will be high

JULY 2005 THE LEADING EDGE 707

 mation. The interpreter categorizes
the seismic data, velocity and den-
sity information, and modeling
and computes a data quality score
which ranges from 0-100%. Most
data quality values are in the range
of 40-85%.
Quite often, more than one seis-
Downloaded 05/30/14 to 130.207.50.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

mic data set is used to interpret an

amplitude anomaly on a prospect;
therefore, the assumption made
here is that the highest quality seis-
mic data are employed to make the
final interpretations. Table 2 lists
the seismic data quality factors. Of
particular significance are the type
of processing/ migration and the
vintage of the seismic data. For
example, the amplitude preserva-
tion of prestack depth-migrated
data (e.g., wave equation) in the
late 1990s is not as good as that
available using today’s algorithms.
In addition, seismic data acquired
and processed for large specula-
tive group seismic surveys quite
often have insufficient acquisition
and processing parameters (migra-
tion velocities) to accurately image
amplitude anomalies for specific
prospect evaluation. Most modern
3D surveys have sufficient bin size
to properly image amplitude
prospects, but 2D data will always
have out-of-plane resolution limi-
tations. Knowledge of the phase of the data (optimally zero
phase) is extremely important for amplitude evaluations with
both prestack and poststack data.
Our industry has numerous types of prestack data and
products; however, essentially all are based on the original
trace gathers and their associated acquisition parameters
and processing. Intercept versus gradient crossplots, fluid
factor displays, or AVO hodograms are all a function of the
trace gather parameters, quality of the processing, and the
distinction of signal from noise on the data.
For rock and fluid property data, the quality and prox-
imity of known density and velocity (P- and S-wave) infor-
mation is extremely important because it relates to the
interpreter’s confidence level of the associated acoustic re-
sponse of the prospective anomaly. The depth, stratigraphy,
and depositional environment of this information in rela-
tion to the amplitude feature have a direct bearing on eval-
uating the prospect. Knowledge of the pressure environment
of the amplitude prospect is also important because the
Figure 5. Location of drilled U.S. prospects in the DHI Consortium data- acoustic response of gas sands in a pressured environment
base. is not the same as in the shallower normal pressured sec-
tion of the sediment column. Finally, modeling and fluid sub-
if the prospect is in a development setting with significant stitution (such as HCI modeling, AVO modeling, or
well control. Initial Pg is a very important anchor point multicomponent modeling) may have an important bear-
because, depending on how the interpreter answers ques- ing on the confidence level in the interpretation of the rock
tions about data quality and amplitude characteristics of the and fluid properties.
anomaly in the following steps, initial Pg will be modified
positively or negatively to get the final or revised Pg. Quantifying the amplitude characteristics. Depending on
the AVO classification (1-4) for a specific prospect, the inter-
Measuring data quality. In order to properly risk seismic preter will answer questions about a multitude of ampli-
amplitude anomalies, it is necessary to evaluate the quality tude characteristics on a scale of 1-5 (worst to best). Each
of the seismic data and nearby relevant rock physics infor- possible response is described and the specific response

708 THE LEADING EDGE JULY 2005

Downloaded 05/30/14 to 130.207.50.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Figure 8. Histograms of the raw data (left) and normalized responses

(right) for the consistency in mapped target area characteristic. The well
outcome is color-coded.

Figure 6. Revised Pg versus data quality score color-coded by the well

outcome.

Figure 9. Histograms of the raw data (left) and normalized responses

(right) for the downdip conformance (fit to closure) based on far offset or
stacked data characteristic. The well outcome is color-coded.

Figure 10. Histograms of the raw data (left) and normalized responses
(right) for the change in AVO compared to model (wet versus hydrocar-
bon-filled) characteristic. The well outcome is color-coded.

Figure 7. The delta Pg (difference between initial Pg and revised Pg) Figure 11. Histograms of the raw data (left) and normalized responses
segregated by successful wells in red and dry holes in blue. (right) for the high porosity clean wet sands pitfall characteristic. The well
outcome is color-coded.
chosen is scored and weighted to determine how Pg will be
revised before the data quality score is included. The con-
sortium has developed an extensive list of pertinent ampli-
tude characteristics; the list ranges up to 36 characteristics
for class 3 anomalies and drops to 25 for class 1 anomalies.
To assist the estimator, the amplitude characteristics are
organized into nine categories: local change in amplitude,
edge effects, rock physics, primary AVO effects, AVO at-
tribute crossplots, amplitude anomaly interpretation pit-
falls, vertical and lateral context, seismic analogs, and
containment and preservation. Figure 3 is an example of
characteristics associated with the first three amplitude cat- Figure 12. Histograms of the raw data (left) and normalized responses
egories (for class 3 anomalies) and related responses by an (right) for the wet reservoir with low gas saturation pitfall characteristic.
interpreter. The well outcome is color-coded.

JULY 2005 THE LEADING EDGE 709

 their associated trends.
Figure 8 displays the responses and trend for the con-
sistency in mapped target area amplitude characteristic
which is in the local change in amplitude category. This char-
acteristic relates to how laterally consistent the amplitude
anomaly appears in the mapped target area on stacked data.
In other words, is the amplitude feature consistently pre-
sent in the mapped closure or does the amplitude appear
Downloaded 05/30/14 to 130.207.50.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

spotty? As the normalized histogram suggests, there is a

strong correlation between the rating of the responses for
this characteristic and the successful outcome of the wells.
One of the most important amplitude characteristics is
the downdip conformance (fit to closure) on far-offset or
stacked data (Figure 9). This characteristic describes how
well the amplitude anomaly correlates to the downdip struc-
ture of the prospect. If the amplitude is truly related to the
structural closure, there should be a good correlation at the
downdip limit. A significant factor in making this interpre-
Figure 13. Geologic chance factor failure modes for the 46 dry holes in the tation is whether the maps are in depth or time, presence
database. of velocity variations, stratigraphic variations, and imaging
quality of the seismic data (e.g., migration).
Database and results of drilled prospects. From 2001 to the A standard approach to interpret AVO features is to
spring of 2004, 18 member oil companies contributed infor- model nearby wells substituting hydrocarbons in wet sands
mation from 93 drilled prospects to help establish a work- and vice versa, in an effort to understand the AVO response
flow and database of amplitude-related prospects used in for prospective reservoirs. Typically Gassmann fluid sub-
the determination of Pg. Members contributed prospects stitution approaches are employed to model the seismic
from a wide variety of prospect types from geologic trends response to fluids in sands. The histogram of Figure 10 dis-
around the world, but in all cases they provided the perti- plays the responses of the change in AVO compared to
nent information necessary to determine the Pg of the ampli- model (wet versus hydrocarbon filled) characteristic. There
tude-related prospect based upon the information that was is a definite positive trend for successful wells with high
known at the time the well was drilled. Of these drilled wells, scores related to this characteristic.
46 were dry holes and 47 were successful. Figure 4 displays In the amplitude anomaly interpretation pitfalls category,
the location of these prospects, which were predominantly one of the many pitfalls identified by the consortium is the
from the United States, but also included prospects in north- possibility that the anomaly was caused by high porosity
west Europe, Far East, West and North Africa. The break- clean wet sands (Figure 11). Depending on the geologic set-
down of the U.S. prospect settings is shown in Figure 5. ting, high porosity clean wet sands can produce amplitudes
The scatter plot of Figure 6 displays the revised Pg, (i.e., that are at times difficult to differentiate from hydrocarbon-
the revised probability of geologic success after the data qual- generated features. Knowledge of sand responses from well
ity and amplitude characteristics have been determined) control in the area, fluid substitution modeling, and local
versus the data quality score. The success (red) or failure depositional models can help determine the appropriate
(blue) of the wells are color coded. Note that there were no response for this characteristic.
dry holes that had a final Pg of over 40%. In fact, the aver- Another important potential pitfall in interpreting seis-
age final Pg for the dry holes was 20%, whereas the aver- mic amplitudes is the possibility the feature was generated
age final Pg for the successful wells was 48%. The data also by a wet reservoir with low gas saturation (5-10% gas). As
indicate that there were no successful wells that had a revised Figure 12 indicates, there is a positive trend of successful
Pg of less than 20%. There is a slight trend suggesting bet- wells with a better anomaly score for this characteristic. The
ter data quality relates to successful wells. interpretation of low gas saturation has long been a prob-
The histograms of Figure 7 display how much the ini- lem in the unconsolidated portion of the sediment column
tial Pg was modified (delta Pg) after the amplitude-related because a small amount of gas can generate prominent
prospects were evaluated for data quality and amplitude reflecting amplitudes. It is very important to understand not
characteristics. This information is further segregated by only the geologic setting, but also the pressure environment
the outcome of the individual wells. The database also indi- in which the prospect is located because the amplitude
cates that any well with a positive delta Pg of approximately response of low gas saturation sands under high pressure
20% or greater was a successful well. By the same token, is different than in the shallow part of the sediment column.
over a third of the dry holes had a negative delta Pg. This Seismic amplitudes generated by hydrocarbons in a high
display relates the confidence level that the amplitude anom- pressure environment tend to not generate dramatic ampli-
aly evaluated truly is generated by hydrocarbons, and the tude increases at low gas saturation levels.
detrimental effect of not having an anomalous event when,
according to empirical observations along trend or model- Failure modes. At the conclusion of the consortium in 2004,
ing efforts, that one should be present. 46 dry holes were in the calibration database. Figure 13
breaks out the failure modes for these dry holes based on
Impact of amplitude characteristics. In an effort to deter- the five geologic chance factors. Closure and containment
mine the significance of the amplitude characteristics to the (seal capacity) were the most significant interpreted causes
success of a well, histograms were generated from the raw for failure with 27 wells (59%). It is at times difficult to dif-
scores for each characteristic, color-coded based on the out- ferentiate structural and stratigraphic closure problems from
come, and normalized to 100% to determine any trends. seal related issues, which is why these two geologic chance
Figures 8-12 show histograms for five characteristics and factors were grouped together. The lack of reservoir qual-

710 THE LEADING EDGE JULY 2005

 ity rock accounted for 14 of the dry holes (30%). Failures al. (TLE, 1998). “Fizz water and low gas-saturated reservoirs”
related to this geologic chance factor included siltstones, con- by Han and Batzle (TLE, 2002). “Rudiments in performance
densed zone, marls, top of hard pressure, and soft shale tracking, with special emphasis on amplitude-bearing
between two “tite” sands. The failures due to migration and prospects” by Roden and Citron (presented at 2004 SEG Annual
source rock factors seen on 5 dry holes (11%) included issues Meeting workshop on Uncertainty in Reservoir Prediction and
such as heavy oil and coal. Reservoir Risk Analysis). TLE

Conclusions and future. Formed in 2001, the DHI Risk Acknowledgments: The authors thank the member companies of the DHI
Downloaded 05/30/14 to 130.207.50.37. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Analysis Consortium has provided a forum to apply the col- Risk Analysis Consortium for providing invaluable information neces-
lective wisdom of dozens of experienced geoscientists from sary to develop the resulting interpretation process and prospect data-
18 different oil companies as well as several consultants, to base. We also thank Gary Citron and Robert Otis of Rose and Associates
the problem of systematic and objective seismic amplitude for review of the manuscript.
prospect evaluations. By the end of 2004, the consortium
had recorded in their archive predrill and postdrill data on Corresponding author: rodenr@wt.net
93 prospects evaluated by this process, a database difficult
to duplicate in the industry. With the consistent application
of this process and continued compilation of additional
prospects, a database of significant size will enable accurate
and definitive statistical information for calibration.
In 2005, the consortium is continuing to add prospects
to the database and is developing a methodology to deter-
mine how seismic amplitudes affect the determination of
productive area and thickness variables in prospect reserve
calculations.

Suggested reading. “Amplitude-versus-offset in gas sands”

by Rutherford and Williams (GEOPHYSICS, 1989). “Framework
for AVO gradient and intercept interpretation” by Castagna et

JULY 2005 THE LEADING EDGE 711