Seismic Reservoir Characterisation of a Channel Sand Oil and Gas Field, Malaysia

T. J. Focht* (Newfield Peninsula Malaysia Inc), M. Sams (IkonScience/ Fugro-Jason (M) Sdn Bhd),
D. Brookes (Newfield Peninsula Malaysia Inc) & J. Ting (Fugro-Jason (M) Sdn Bhd)

Introduction
A regional, straight sandstone channel of mid-Miocene age within the Malay Basin forms a reservoir
at several areas where the flow direction abruptly turned forming large meanders. One such meander
constitutes the study field where oil and gas have been stratigraphically trapped. The seismic response
to this channel is complicated by the laterally varying fluid type. The channel when wet gives a Class
II AVO response and Class III when sufficient thickness of gas is present. The response changes
depending on the height of the channel above the gas and oil contacts. The oil leg is only 12 meters
thick. The response is also complicated by smaller channels just above and below this large channel.
An underlying coal that varies in thickness and in some places appears to have been cut by the large
channel also can distort imaging. With three exploration wells available, a seismic inversion study
was conducted to try to extract more reliable and quantitative information from the seismic. As a
result of the inversion the interpretation of the channel has been modified (Figure 1) and the inversion
predictions have so far been confirmed by an additional two appraisal and four development wells.

Figure 1 The channel distribution with the predicted fluids prior to the inversion study (left) and after
the inversion study (right). Gas is shown in red and oil in green. The study field is circled. The line on
the right represents the line of cross-section shown in Figure 3.
Geology
The sole reservoir of the field is a fluvial channel sand and the hydrocarbons are stratigraphically
trapped. The channel sand was deposited in an incised valley cutting into upper-coastal plain deposits
consisting of siltstone and coals. All of the vertical wells to date have encountered a sandstone with a
consistent thickness of 30 meters. A conventional core was acquired in one well which shows a
fining-upwards grain size and a gravel lag at the base. The sandstone in the core is of good quality.
Porosity ranges from 25 to 27 pu and net-to-gross exceeds 90 percent. Petrophysical analysis of all the
wells in the field indicates that the reservoir shares common gas-oil and oil-water contacts.
Rock physics
At the time of the seismic characterisation study there were three wells drilled within the channel
(Figure 1). One was wet, drilled on an amplitude anomaly down dip to the south. Two encountered
gas, oil, and water. All three penetrated approximately 30 meters of good quality, channel sandstone.
The logging suites in all of the wells were good quality with shear sonic measured in the wet well and
one of the hydrocarbon bearing wells. At this time core data was not available. After further drilling
Petroleum Geoscience Conference & Exhibition
Delivering Value: Realising Exploration & Development Potential
Kuala Lumpur, Malaysia, 23-24 April 2012
with one appraisal well cored, it was found that the clay content of the channel sand was around the
25% level and the porosities were in the 25 to 27 pu range. The oil encountered has an API of 39.
Cross-plotting of the P-impedance and Vp/Vs logs showed that there was separation such that at log
resolution gas, oil and brine sands and shales could be differentiated. A rock physics model was
developed that could predict to a reasonably high degree of accuracy the elastic logs based on a
petrophysical interpretation in terms of volume of clay, porosity and saturation. During this modelling
it was apparent that the elastic logs were affected by invasion and corrections were applied. Shear
sonic data were predicted for the one well with no measured shear, and various fluid scenarios were
modelled to try to understand the seismic response to various scenarios. Forward modelling of the
AVO response indicated that the seismic response would be Class II to Class III and the ability to use
high angle data would greatly enhance the ability to identify the channel and fluid fill. The rock
physics modelled logs were used for tying the wells to the seismic and extracting wavelets for
inversion.
Seismic conditioning
The seismic data available at the time had been stacked into 3 angle sub-stacks between 5 and 40
degrees, which initially seemed to be sufficient for the purpose of extracting useful Vp/Vs information
from seismic inversion. However, a look at the NMO corrected seismic gathers showed that the events
were not flat across the entire offset range and deviated upwards significantly at far offsets. The far
traces appeared to contain useful amplitude information. It was decided to re-pick the seismic
velocities and include the effects of anisotropy to flatten the gathers (Figure 2) and then re-stack to 5
angle stacks between 5 and 55 degrees. Validation for this process came with the drilling of a highly
deviated well, which showed that the shales within the section above the channel are indeed highly
anisotropic. After removing the residual moveout the seismic gathers were further processed to reduce
random noise and multiples. The removal of multiples has had a significant impact in terms of
cleaning up the image of the channel. A cross cutting channel depicted in Figure 1 in the pre-inversion
channel distribution was found to be a multiple.
Despite this conditioning there remained a significant acquisition footprint. This footprint was
assessed through RMS attributes at various intervals. These attributes were used to scale each of the
seismic sub-stacks individually. Quality control checks were made to ensure that this process did not
introduce artificial variations in the inversion results.

Figure 2 Seismic gathers before (left) and after (right) correction for anisotropic moveout. The
channel is the event at the centre of the time interval and is clearly a Class II AVO. After anisotropic
moveout correction further conditioning was applied.
Inversion
Seismic inversion is run to convert seismic reflectivity to elastic properties. Due to a lack of low
frequency information in the seismic data a low frequency model is required. This model is integrated
during the inversion to provide a broadband result. When a number of angle stacks are inverted
simultaneously the results are in terms of P- and S-impedance (or combinations of these) and density,
though the later is most often not well constrained. This means that low frequency models for P- and
S-impedance are required. These low frequency models must be geologically reasonable and can
often only be determined through the interpretation of intermediate inversion results. For example, in
this case all the well data penetrate the channel sand with various thicknesses of hydrocarbons.
Petroleum Geoscience Conference & Exhibition
Delivering Value: Realising Exploration & Development Potential
Kuala Lumpur, Malaysia, 23-24 April 2012
Clearly the elastic properties of the sandstone cannot be extrapolated beyond the channel limits across
the entire area to form a low frequency model. To overcome this, a detailed model of the sandstone
reservoir and encasing strata was made and updated with intermediate inversion results. The detailed
model is filtered back to 8 Hz to provide the low frequency component for the inversions and filtered
back to 60 Hz to provide a comparison with the inversion output. Differences between the model and
inversion are used to update the model.
The model is primarily based on an interpolation of the wells after the channel properties have been
removed and replaced by a linear interpolation. Outside of the channel there is a high degree of
correlation between the wells that justifies this procedure. The distribution of the channel is initially
determined from an inversion of the ultra-far angle stack using a simple trend as a low frequency
model. This proves to give a good prediction of the top and base of the channel at the well locations.
Within the interpreted channel the porosity from the wells is interpolated. The porosity is then
transformed to elastic properties based on the rock physics model and the fluid contacts. The fluid
contacts are placed in time based on a time to depth conversion as described next.
Time to depth conversion
Time to depth conversion is important in this project. The structural dips are low and therefore small
variations in time to depth prediction can have impact to the structural maps. The development plan
that includes drilling long horizontal wells within the 12 meter oil column also requires accurate time
to depth conversion.
Despite the apparent lateral consistency of the elastic properties above the channel, the time to depth
relationships at the wells indicated that a simple model was not sufficient. The seismic velocities from
the re-picking were conditioned and calibrated to the wells to provide a velocity model with which to
convert the top of the channel to depth. This top channel in time and depth can then be used as a
datum and velocities from the detailed low frequency modeling scheme used to time-depth convert
data at other levels.
Interpretation
Interpretation is an iterative procedure in a seismic inversion workflow. There are a couple of critical
items that are required from the interpretation. First is the ability to identify and pick the top and base
of the channel. It is important to know where the channel exists and whether it has the same reservoir
quality as penetrated by the wells. The top and base were interpreted on a number of inversion results,
based on forward modelling studies that highlight the impact of fluid fill on how to pick the top (and
base) of the sand. The inversion of the very far stack and the Vp/Vs from the simultaneous inversion
of all the angle stacks, proved most useful. The P-impedance on its own is not helpful in determining
the presence of channel sand (Figure 3).
Secondly, reservoir quality and fluid prediction are required. Due to uncertainties in the time to depth
relationships across the field, it is important to get confirmation of the distribution of the fluids.
Bayesian analysis was applied to generate fluid and lithology probabilities from the P- and S-
impedance from the inversion. Due to the resolution and seismic noise these can not give detailed
interpretations but it was possible to extract the most likely fluid at the top of the reservoir for
comparison with that derived from the time to depth conversion of the top channel.
The P- and S-impedance from the inversion are affected by the presence of fluids and in combination
with the limited resolution make identifying variations in porosity (or clay content) rather difficult.
However, since a detailed model of the reservoir has been built to support the low frequency model, it
is possible to look at the differences between this model filtered back to the seismic bandwidth and
the seismic inversion results. These differences were interpreted in terms of variations of porosity
from that initial assumption of interpolating the well porosities.

Petroleum Geoscience Conference & Exhibition

Delivering Value: Realising Exploration & Development Potential
Kuala Lumpur, Malaysia, 23-24 April 2012
Early development drilling
Since the inversion study was completed, another six wells have been drilled. The first was an
appraisal well drilled within the main channel at a location with a significant gas expression in the
seismic. The well came in as prognosed in terms of depth, channel thickness and fluid content. The
second well was also an appraisal well, but this time drilled outside of the main channel. Prior to the
inversion study there was no clear idea of what sub facies of the channel system to expect at this
location. The inversion study indicated an oil bearing channel sand of similar thickness to the other
wells. The well penetrated a 30 meter channel sand with porosities similar to the other wells. No gas
was encountered and the same oil-water contact as seen in the other wells was observed extending the
size of the field. However, the well came in structurally low limiting some of the development
potential.

Figure 3 A cross-section through two recently drilled wells. The direction is south to north along the
line indicated in Figure 1. The left hand panel shows the bandpass inversion results of the ultra-far
angle stack that has been used for the initial interpretation of the top and base of the channel. On the
right is the P-impedance from a simultaneous inversion. In both cases the data have been flattened to
the top of the channel and blues represent low impedances. These wells were drilled after the
inversion study and both came in as predicted in terms of channel thickness and fluid content.
The next four wells were horizontals with 1.5 kilometer lengths within the oil leg. The wells
encountered the top of structure giving four more depth points. Two of the horizontal wells exited the
top of structure at the heel providing two additional depth points. All of these data points were close
to prognosis. Therefore if we trust the pick of the top of the reservoir in time we have more velocity
control points for interpolation. Of course, there will be uncertainties in the top channel pick as it
relies to a large extent on the far angle stacks, whose moveout correction is determined by a
reasonably sparse velocity picking which included anisotropy corrections.
Conclusion
Seismic inversion can be used to improve our understanding and interpretation of seismic data. In
fact, seismic inversion should be considered an integral part of seismic interpretation, where both
mutually benefit the other. Naturally, interpretation is also based on both geological understanding
and rock physics models. Rock physics scenario modelling provides the framework for interpreting
the elastic properties from the seismic inversion. In this case, interpretation has improved as different
inversions have been applied and the inversion results have become more quantitative as the
interpretation has improved. The results for this channel sand are semi-quantitative allowing for the
prediction of the channel thickness and the fluid at the top of the channel. Estimates of porosity
variation have been made but resolution issues and the small variations in the sand porosity make
validation difficult.
Acknowledgements
We would like to thank the management of Newfield and PETRONAS for permission to present these
data.

Petroleum Geoscience Conference & Exhibition

Delivering Value: Realising Exploration & Development Potential
Kuala Lumpur, Malaysia, 23-24 April 2012