You are on page 1of 9

SPECIAL

R o c k SECTION:
p h y s i R
c s
o c k

P h y s i c s
p

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Rock physics estimation of cement volume, sorting, and net-to-gross


in North Sea sandstones
PER AVSETH, Odin Petroleum
ARILD JRSTAD, Lundin-Norway
AART-JAN VAN WIJNGAARDEN, StatoilHydro
GARY MAVKO, Stanford University

ock physics represents the link between geologic reservoir


parameters (e.g., porosity, clay content, sorting, lithology,
saturation) and seismic properties (e.g., VP/VS, density, elastic
moduli). Rock physics models can either be used to interpret
observed sonic and seismic velocities in terms of reservoir
parameters, or they can be used to extrapolate beyond
observed range to predict certain what if scenarios in terms
of uid or lithology substitution. Rock physics models can
also be used to estimate expected seismic properties from
observed reservoir properties.
In this study, we show how depositional and compactional trends can be decomposed and quantied into rock physics
crossplots, and then depicted as new informative well logs.
The estimation of cement volume, sorting, and uid factor as
additional rock physics logs provides geo-practitioners with
useful extra information about reservoir properties not easily
achieved from conventional petrophysical logs. Furthermore,
the logs can easily be used in conjunction with quantitative
seismic interpretation during feasibility or inversion studies.
We also suggest an approach to model and quantify the eect
of net-to-gross and heterogeneity of interbedded sand-shale
sequences on seismic properties.
We demonstrate our integrated rock physics techniques
for a North Sea case where two adjacent wells have very different seismic signatures due to local dierences in diagenetic
and depositional characters. Without quantifying and understanding these geologic dierences, we would not be able to
forecast the correct pore uid in these two neighboring hydrocarbon discoveries. In this paper we concentrate on North
Sea turbidite sandstones of Paleocene age. This complements
the 2008 paper by Avseth et al. focused on the rock physics of
the background shale.
The rock physics ambiguities between pore uids, rock
texture, and net-to-gross
Pore uid sensitivity in reservoir sandstones is highly aected
by reservoir heterogeneity and sandstone microstructure, and
it is therefore important to include these geological factors in
the rock physics analysis (Figure 1). For instance, the onset
of diagenetic quartz cementation of sandstones tends to occur at burial depths corresponding to approximately 70C
(Bjorlykke and Egeberg, 1993). This initial cement drastically changes the AVO signatures of the reservoir and also
reduces the pressure and uid sensitivity of the sandstones
(Avseth et al., 2008). Figure 2 shows schematically the outline of a rock physics template (degaard and Avseth, 2004),
acoustic impedance versus VP/VS, where calibrated rock physics models have been selected that t local data observations
98

The Leading Edge

January 2009

Figure 1. Rock physics represents the link between geologic parameters


and geophysical measurements. Rock texture can be diagnosed from
rock physics models of elastic moduli or velocities versus porosity. The
various models include the unconsolidated sand model and contact
cement model (Dvorkin and Nur, 1996) together with the constant
cement model (Avseth et al., 2000) and are utilized in this article for
quantication of cement volume and sorting (adapted from Avseth et
al., 2005).

(well-log data or seismic inversion data) of various lithologies


and pore uids. If we assume that the models were generated
for unconsolidated and homogeneous sands and shales, the
selected template can serve well as an interpretation tool to
discriminate between the dierent observed categories, and
also to foresee the eect of local changes in porosity (sorting) and saturation. However, what if there are some geologic changes occurring that have not been included in the
selected template? For instance, the presence of diagenetic
quartz cement will move brine-saturated sandstone data in
a crossplot of AI versus VP/VS to an area of very low VP/VS
where we may expect hydrocarbon-saturated sandstones to
plot. On the contrary, reservoir heterogeneity and decreasing net-to-gross associated with interbedded sands and shales
tend to move data points in the direction of the shale cluster.
The cement eect is a microstructural eect, whereas the netto-gross is a scale eect. In the case where the interbedded
shale is relatively soft compared to the sand, the net-to-gross
eect will counteract the eect of cement on eective rock
stiness, hence the opposite direction in the AI-VP/VS cross-

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

Figure 2. (a) Sands lled with oil are normally well separated from
brine-lled sands in a rock physics crossplot of VP/VS versus acoustic
impedance (AI). (b) The eect of initial cement will reduce the uid
sensitivity of sandstones and move brine-sand data toward the oil-sand
model, whereas the eect of net-to-gross will normally move data in the
opposite direction in a VP/VS-AI plot. Hence, oil sands with relatively
low net-to-gross can have higher VP/VS ratios than homogeneous brine
sands with initial cement.

plot. The schematic, but qualitatively correct, illustration in


Figure 2 demonstrates why it is important to include these
geologic factors when analyzing the rock physics properties
and seismic uid sensitivity in reservoir sandstones.
In this paper, we study the eect of rock texture and netto-gross in the Heimdal Formation sandstones in the North
Sea, with focus on two neighboring wells, and show how
variation in these geologic parameters drastically distorts the
uid sensitivity to hydrocarbons. Figure 3 shows a seismic
line with the two wells indicated as black lines. One well penetrated a thick, turbiditic gas sand with a thin oil leg, whereas
the adjacent well penetrated a turbidite sand lled with oil.
However, it turned out that the AVO signatures of the oil
sand is a class III AVO response, whereas the gas discovery
shows a class II response (see AVO crossplot in Figure 3). This
is also documented in the well-log data and crossplot in Figure 4. The gas-saturated top Heimdal sands in well 1 show a
small increase in acoustic impedance, while the oil-saturated
sands in well 2 show a signicant drop in acoustic impedance. In the following, we will demonstrate the importance
of rock texture, in particular cement volume and net-to-gross,

Figure 3. (a) Near, far, and far-near stack sections intersecting two
North Sea hydrocarbons discoveries, and (b) AVO crossplot of far-near
versus near for selected windows around the target interval. The blue
cluster represents a shaly background trend, whereas the oil discovery
plots as a class III AVO anomaly (red) and the gas discovery plots
as a class II AVO anomaly (yellow). These observations are counterintuitive if the sands are assumed to have the same dry-rock properties.
Hence, there must be some geologic factors that are signicantly
dierent for the two neighboring discoveries.

and how these geologic factors aect the seismic signatures of


hydrocarbons within these sands.
Rock physics estimation of cement volume and sorting
Avseth et al. (2008) demonstrated the interplay between
shale and sandstone diagenesis and how the dierent compaction trends aected the seismic signatures in the North
Sea deep-marine sedimentary rocks of Tertiary age. At about
70C, quartz cementation of quartz-rich sandstones initiates
in association with smectite-to-illite transition in embedding
shales. In the North Sea, this corresponds to a burial depth
of about 2 km, and this is around the target depth of the
prolic Palaeocene and Eocene reservoir sands that represent
major prospects for the oil industry. Many geologic factors
inuence quartz cementation, and an excellent overview of
this complex geologic process is given by Worden and Morad (2000). In this paper, we will not debate the controlling
factors, but rather investigate the signicance of quartz cement related to hydrocarbon prediction from seismic data.
We want to nd out if we can observe the transition zone
from unconsolidated-to-cemented sandstones in terms of
January 2009

The Leading Edge

99

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

Figure 5. Shear-wave velocity log data versus total porosity and


superimposed diagnostic rock physics models. Using the models, we
can quantify the cement volume (a) and degree of sorting (b). (Green
points are shale data with high GR values, and are for practical
reasons given the value -1 in cement volume and 0 in sorting).

Figure 4. Log data from (a) well 1 and (b) well 2 in Figure 3, and
(c) plot of AI versus VP/VS for dierent facies and uids in the target
zone compared with rock physics models. The reservoir sandstones
are part of the Heimdal Formation, capped by shales. Note the slight
increase in acoustic impedance at the top of the Heimdal in well 1,
whereas the same interval shows a drop in acoustic impedance in
well 2. In (c), cluster I is the caprock shale in both wells, whereas II
comprises the brine sandstones in both wells. III and IV are reservoir
sandstones in well 1 lled with oil and gas, respectively. V is the upper
oil zone in well 2. Hence, the well-log observations match what we
observe in the seismic, with a class II AVO for the gas zone in well 1,
and a class III AVO for the oil zone in well 2.

rock physics properties, if we can document the initial quartz


cement from thin-section analysis, and if we are able to estimate the cement volume from rock physics relations.
Using the diagnostic models in Figure 1, we can infer
the microstructure from velocity-porosity data (Dvorkin and
Nur, 1996; Avseth et al., 2000). With good local validation
of the models, we can even quantify the degree of sorting and
100

The Leading Edge

January 2009

cement volume from these diagnostic crossplots (Figure 5).


We choose to do this in the VS versus porosity domain, in
order to avoid pore uid eects. (Shear modulus would be the
optimal elastic parameter for this exercise, theoretically with
zero uid eects. However, the estimation of shear modulus
includes VP/VS and density and the uncertainties associated
with varying mud-ltrate invasion can deteriorate the results
in the hydrocarbon-saturated intervals.) The cement volume
is estimated by interpolating between the constant cement
volume trends, whereas the sorting is dened by the observed
porosity () normalized by the high end-member porosity
along the given constant cement trend, that is the porosity
connecting the constant cement model with the contact cement model.
Having estimated cement volume and sorting, we can
plot these as logs and compare with other petrophysical logs.
Figure 6 shows the resulting estimation of cement volume
and sorting. The middle subplot shows cement volume as
magnitude and sorting as superimposed color, whereas the
right subplot shows sorting as magnitude and cement volume

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

Figure 6. Estimated cement volume and sorting versus depth for


well 1. Note the onset of cement starting at around 2000-m depth
(corresponding to 70C) and the increasing cement volume with depth
(second column). Sorting, however, shows a more erratic pattern with
no depth trend (third column).

Figure 7. Thin sections from Heimdal sands. The upper image (a)
shows loosely packed, poorly consolidated sands (scale bar equals 0.5
mm). Analysis of a zoomed-in image (b) conrms the presence of
quartz overgrowths and contact cement. On detrital quartz grains,
we observe dust rims representing the original grain surfaces that have
been covered by quartz cement (arrows). Feldspar overgrowth and
calcite cement also occur, yet quartz cement is dominating.

102

The Leading Edge

January 2009

Figure 8. The estimated cement volume based on rock physics models


(black line in right plot) compared with point-count cement volume
in the upper 100 m of the Heimdal reservoir sands in well 1. The
estimated cement volume is slightly larger than the point-counted
quartz cement, but for most of the depth-range it matches nicely with
the point-counted total cement volume.

as color. For the relatively clean Heimdal sandstones starting


at around 2 km, we observe a clear depth trend in the cement volume. This matches with observations made by, for
instance, Giles et al. (2000) and Walderhaug et al. (2000).
The sorting, however, shows a more erratic pattern, without
a depth trend, which is expected since sorting is associated
with depositional trends. (Also note a small cement volume
increase around 1800 m, which is within the Balder Formations tuaceous interval; amorphous silica in the tu is
known to be a source of quartz cement.)
It is essential to verify the presence of initial cementation
predicted from the rock physics relations with thin-section
observations from the reservoir sands penetrated by this well.
Figure 7 shows a thin section from the relatively clean Heimdal sands but, at rst glance, the sandstone looks unconsolidated with grains loosely arranged and moderately well
sorted. A closer investigation, however, reveals the presence
of initial quartz overgrowth covering original grain surfaces
indicated by dust rims (arrows in Figure 7b). This observation conrms what we see in the rock physics crossplots of
the well-log data. It is interesting that the well-log data with
10 s of cm resolution reect what we observe at the microscale.
Figure 8 compares the estimated cement volume and the cement volume from the thin-section point count analysis (only
available in the upper part of the Heimdal sands in well 1).
The volume of quartz cement from the thin-section analysis
is somewhat lower than the volume estimated from the rock
physics models. However, if we compare with the total cement volume, the match is very good. There are other types of
cement than quartz, like feldspar overgrowths and carbonate
cement reported for Heimdal Formation sandstones. Most of
the cement, however, is quartz.
It is interesting to compare our new textural logs with logs
that reect pore uid content. Avseth et al. (2008) presented
a way to estimate the uid-factor log on a sample-by-sample
basis for well-log data, using a smooth depth-dependent shale

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

zone is correlating with a uid-factor anomaly. This could be


a calcite bench that is characterized by high cement volume at
the same time as it gives a large deviation from the mudrock
line, and therefore a uid factor. The oil zone shows a very
weak uid factor, and this may be because of the heterogeneous nature of the oil zone (i.e., more laminated). In the
adjacent well 2, we observe a uid factor only in the upper
part of the oil zone. A closer look at the predicted cement
volume for this well shows that the cement volume starts to
increase within the oil zone. Probably, the presence of clay
and lamination in the upper part of the oil reservoir in well 2
has inhibited the quartz cementation.
Increased heterogeneity and related decreased net-to-gross
in turbidite sands often imply that the interbedded sands
have higher clay content than more massive facies associations (i.e., thick-bedded turbidite sands), and therefore thinbedded sands tend to have lower quartz cement volume. The
presence of oil can also inhibit quartz cementation, especially
if the reservoir is oil-wet (Barclay and Worden, 2000).

Figure 9. Estimated uid-factor log (colored logs in third plot) and


cement volume (colored logs in fourth plot) for well 1 (a) and well
2 (b), respectively. Juxtaposed logs include shale volume and porosity
(left plots), hydrocarbon saturations for well 1, second plot in (a), and
deep resistivity for well 2, second plot in (b). Superimposed on the
uid-factor log is the dierence between predicted shear-wave velocity
(using Greenberg-Castagna relationship) and observed shear-wave
velocity, whereas shear modulus is superimposed onto the estimated
cement volume logs, showing strong correlations for both. Note that
the uppermost oil-lled Heimdal sands in well 2 have lower cement
volume than the uppermost gas-lled Heimdal sands in well 1.

model as the background trend. By juxtaposing the estimated


cement volume with the uid factor, we obtain complementary information about expected seismic signatures. Moreover, it can be interesting to see if the two logs are somehow
correlated or completely independent. Figure 9 (third upper
plot) shows the uid-factor log in color for the target zone of
well 1. Superimposed on the uid-factor log is a black log representing the dierence between predicted VS (local mudrock
line, modied from Castagna et al., 1985, with observed VP
as input) and observed VS. These are clearly correlated, since
uid factor is dened as deviation from background trend in
a VP/VS crossplotthat is the mudrock line. Also note how
the uid-factor log is nicely correlated with the saturation
logs in the second subplot. In particular, the gas zone (yellow
saturation curve) gives a strong uid factor. The oil zone (red
saturation curve), however, shows a very weak uid factor. In
the fourth subplot, we include the cement volume together
with the shear modulus. There is clearly a strong correlation
between cement volume and shear modulus, as expected.
Also note a peak in cement volume below the hydrocarbon

Rock physics estimation of net-to-gross


There is increasing focus on hydrocarbon prediction from
heterogeneous reservoirs. In this paper, we have already indicated that the Heimdal Formation turbiditic sandstones can
be rather heterogeneous. One useful parameter for quantication of the heterogeneity of sands is net-to-gross (N/G),
that is the fraction of clean, permeable sand to the complete
reservoir including reservoir sands and intercalating impermeable shales. However, N/G is a scale-dependent parameter. A reservoir zone can have high N/G at well-log scale,
but low N/G at seismic scale. Nevertheless, N/G is a useful parameter when we upscale from alternating thin beds
of dierent lithologies and/or uid saturations to an eective medium, during rock physics analysis of well-log and
seismic data. It is also a parameter that geologists are very
familiar with. Takahashi (2000) formulated how to predict sand-shale ratio based on statistical rock physics simulations of various bedding scenarios. Vernik et al. (2002)
predicted N/G from P and S impedance inversion results.
Stovas et al. (2006) used eective medium theory combined
with Gassmann theory to predict N/G and saturation from
AVO attributes. Finally, Connolly and Kemper (2007) used
a statistical and data-driven approach to predict what they
referred to as net pay from turbiditic reservoirs oshore Angola. However, little eort has been devoted to quantify N/G
from well-log and seismic data using rock physics models. In
this paper, we present a ve-step methodology to model the
rock physics properties of interbedded sands and shales with
dierent pore-uid saturation scenarios (Figure 10; also see
Avseth et al., 2006).
Step 1 is the estimation of dry bulk and shear moduli,
Kdry and +dry at critical porosity (Figure 10, upper left) using Hertz-Mindlin contact theory. In step 2 (Figure 10, lower
left), we interpolate between the high-porosity end member
and the zero-porosity mineral point, choosing either the
modied lower or upper Hashin-Shtrikman bound. These
two bounds have been found to nicely predict the elastic
January 2009

The Leading Edge

103

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

Figure 10. Illustration of the ve-step approach for modeling the


eective moduli of interbedded sands-shales with varying net-to-gross.
First estimate the elastic moduli for the sand layers at the highporosity end-member using Hertz-Mindlin contact theory. Next use
Hashin-Shtrikman models to interpolate between high-porosity end
member and mineral point. Gassmann uid substitution for dierent
saturations is performed for all porosity values. Finally, Backus
averaging is performed to estimate the eective rock properties of the
intercalated sand-shale sequences, where hydrocarbons are only located
within the sandy layers.

moduli as function of porosity reduction for unconsolidated


or cemented rocks, respectively (Avseth et al., 2005). In step
3 (Figure 10, upper right), we apply Gassmann theory and
perform uid substitution and estimate elastic properties of
clean sands with uniform, but varying saturation of brine
(Sw) and hydrocarbons (1-Sw), for all porosities. In step 4
(Figure 10, lower right), we select a characteristic shale to be
interbedded with the clean sand units. We can either derive
typical shale properties from well-log data in an area of study,
or we can repeat steps 1 and 2 to simulate the properties of
shale. However, we assume that the shale will be completely
impermeable to hydrocarbons, and that the shale layers will
only be saturated with brine trapped during deposition. It is
reasonable to assume that the porosity of thin-bedded shales
at a given depth will be fairly constant, in contrary to the
porosity of the thin-bedded sands that is prone to vary with
sorting. Hence, we assume a constant characteristic porosity
for shale, qsh. In step 5 (Figure 10, lower right), we apply
Backus average eective medium theory to estimate the eective, upscaled anisotropic properties of the interbedded shalesand sequences. We do this for various net-to-gross values,
ranging from 0 to 1. The Backus average approximates a stack
of alternating thin layers of two isotropic media as one eective anisotropic medium. This medium will be characterized
by ve independent elastic moduli according to the transverse
isotropic elasticity matrix (see appendix). The various elastic
moduli that we need further are found from the velocities,
densities, and the fractions of the alternating sand and shale
layers. Finally, we can derive diagnostic rock physics models
for varying net-to-gross.
We apply the ve-step rock physics methodology outlined
above to the interpretation of well-log data from the turbiditic North Sea reservoir in well 1. The well-log data in Figure 4
include the gamma-ray log, and we observe that the turbidite
sands are fairly massive, but thin beds of shales are clearly in104

The Leading Edge

January 2009

Figure 11. Rock physics models of acoustic impedance (AI) versus


VP/VS, including models for varying net-to-gross (N/G) and gas
saturation (1-Sw), created by the ve-step procedure outlined
in this paper. (a) The brine sands (cyan) and oil sands (red) fall
between N/G=1 and N/G=0.8. Most gas sands (yellow) seem to
fall on similar porosities, with N/G varying between 1 and 0.6.
Studying the gas sands (yellow) in detail (b), we clearly see that the
sands span VP/VS ratios from 1.5-1.9. Most gas sands have VP/VS of
1.61.7 representative for N/G values of 0.9. Hence, very little shale
intercalation will cause a signicant increase in VP/VS compared to
homogeneous, clean sands (N/G=1).

tercalated between sand units. We create a crossplot of acoustic impedance and VP/VS data from the target zone based on
the log data from well 1. This crossplot is compared to rock
physics models for various net-to-gross ratios made according
to the ve-step methodology above. The characteristic shale is
picked from the caprock shale above the reservoir. Note that
this may be somewhat erroneous because the caprock shale
is not necessarily equivalent to the interbedded shale within
the reservoir. In Figure 11, we can see the various models for
net-to-gross (N/G) of 1, 0.9, 0.8, and 0.6. For each N/G, we
have included varying gas saturation within the sand layers.
It is interesting to note how the decrease in N/G will cause a
drastic increase of VP/VS, regardless of porosity, even for high
gas saturation in the sands. The acoustic impedances drop
drastically when the sands have low porosities, but increase
slightly when the sands have high porosity. This is of course
due to the relative contrast to the intercalating shale. The res-

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

Figure 12. Core images from well 1 juxtaposed with well-log data (GR, AI, and
VP/VS). All are taken from the gas-saturated zone of Heimdal Formation sandstones. Note
that VP/VS in the homogeneous zones (a) are signicantly higher (1.61.7) than in the
heterogeneous zones (b) (1.81.9), in accordance with the results from the net-to-gross
modeling (Figure 11).

ervoir sand data points match nicely with the


models for high N/G values (0.81.0), and at
the rst glance it looks like the gas sands fall
close to the line for homogeneous (N/G = 1)
sandstone with 100% gas.
In Figure 11 (lower), however, we have
removed the brine-lled and oil-lled sand
data, leaving only the gas-saturated sands
together with the cap-rock shales. Now, we
clearly observe that the gas-saturated sands
span a wide range of VP/VS ratios. Some gas
sands seem to have N/G values between 0.8
and 0.6, causing VP/VS close to 2, a value
more typical of brine-saturated sands. This
observation is consistent with a patchy saturation behavior. The heterogeneities of the intercalating shales are causing a geologic control on the saturation pattern. Even though
the gas saturation is uniform at the scale of
the sandstone porosity, the eective medium
will experience patchy saturation when the
sand-shale layers are much smaller than the
resolution of the sonic or seismic waves. This
is further documented in Figure 12 where we
juxtapose core images from the gas zone in
well 1 with logs of AI and VP/VS. In the very
homogenous zone (Figure 12a), VP/VS ratios
are between 1.61.7. In the more laminated
zone (Figure 12b), VP/VS ratios are between
1.81.9.
Hydrocarbons versus lithology from seismic data
Avseth et al. (2008) showed how to obtain
the optimal uid factor based on rock physics modeling of a shale compaction trend. In
this study we focus on the sandstone reservoir
and, in addition to the uid-factor attribute,
we derive the dierence between intercept
and gradient, which has been shown to be

Figure 13. (a) AVO crossplots with uid factor


and VS/VS as colored attributes, for the seismic
section intersecting the two wells in this study. The
uid-factor attribute represents the deviation from
local background shale trend. The VS/VS attribute
reects rock stiness, and it is expected to be an
appropriate attribute to detect initial quartz cement
in the Heimdal Formation sandstones. The resulting
AVO attributes (b) show strong uid-factor values
at both wells, whereas the VS/VS attribute shows
a much stronger contrast at well 1 than at well 2.
This is likely reecting what we have observed in the
well-log data, that the uppermost part of Heimdal
sandstones in well 2 are less cemented than the
Heimdal sandstones in well 1, and this can also
explain the dierent AVO classes at the two well
locations.
106

The Leading Edge

January 2009

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

close to VS/VS (Fatti et al., 1996). This attribute should not


be signicantly aected by pore uids, but very sensitive to
lithology and cementation. Figure 13 shows the uid-factor
line juxtaposed with the VS/VS attribute for the seismic line
intersecting both well 1 and well 2. Lo and behold, both discoveries show strong uid factors, whereas only well 1 shows
a strong change in VS/VS at the top of the reservoir. This
matches with the observations from the well-log data, and
the rock physics estimation of relatively low cement volume
at the top of the Heimdal sands in well 2. It further explains
why we have such drastically dierent AVO signatures in
the two wells, and the counter-intuitive AVO classes for oil
and gas are explained by the local change in rock texture.
Decreasing net-to-gross will likely give similar eect on the
seismic data as decreasing cement volume, since these eects
have similar trends in AI-VP/VS crossplots (Figure 2b). In fact
the two eects can superimpose each other and, as discussed
earlier, we expect the lower quartz cement volume at the top
of the Heimdal sandstones in well 2 to be associated with
lower net-to-gross, compared to what is observed for well
1. With two seismic parameters, R(0) and G, it is possible
to discriminate the uid eects from the lithologic eects,
but it can be dicult to discriminate the cement eect from
the net-to-gross eect using a deterministic rock physics link
between geologic parameters and seismic properties. For improved seismic characterization of the reservoir geology into
dierent uids, textures, facies, and bedding congurations,
we suggest the use of statistical rock physics and probabilistic
classication of the AVO data.

Appendix: Estimating elastic properties of interbedded


sand-shale intervals
The Backus average approximates a stack of alternating thin
layers of two isotropic media as one eective anisotropic medium (Mavko et al., 1998). This medium will be characterized by ve independent elastic moduli according to the following elasticity matrix:

where

The various elastic moduli that we need further are found


from the velocities, densities, and the fractions of the alternating sand and shale layers:

,
Conclusions
Initial quartz cementation, sorting, and varying net-to-gross
are all important geologic factors that aect the seismic uid
predictability in reservoir sandstones. Both reduced net-togross and increasing cementation will reduce uid sensitivity
of the reservoir sandstones. Net-to-gross will do so by nonuniform (patchy) saturation and cement by increased dry
rock stiness. Deteriorating sorting tends to increase uid
sensitivity due to lower porosity without signicant stiening of rock frame. These geologic factors are important to
consider during uid prediction from seismic data, especially
when the target depth is around the depth/temperature of
initial quartz cement (circa 2 km/70C). A case example
from the North Sea demonstrates how the geologic factors
are aecting the seismic signatures, and causing counter-intuitive AVO anomalies for two adjacent hydrocarbon discoveries. Through quantitative rock physics links, we are able to
estimate cement volume and sorting from well-log data, and
it is appealing to see how quartz cement conrmed at the microscale (i.e., thin-sections) aects the macroscale (i.e., sonic
and seismic signatures) of the studied reservoir sands. Still,
the rock physics properties are changing from one scale to
the other due to heterogeneities (i.e., interbedding), and we
have presented an approach to quantify the eect of net-togross on seismic properties. Ultimately, we are able to explain
the observed sonic and seismic signatures in terms of uid
and rock properties.

From these moduli we can express vertical and horizontal


seismic velocities propagating through the eective medium:
,

,
The average density of the eective medium is found by the
arithmetic average of densities of the alternating lithologies:

Suggested reading. Elasticity of high-porosity sandstones: Theory for two North Sea data sets by Dvorkin and Nur (Geophysics, 1996). Rock rhysics diagnostics of North Sea sands: The
link between microstructure and seismic properties by Avseth et
al. (Geophysical Research Letters, 2000). Quantitative Seismic

InterpretationApplying rock physics tools to reduce interpretation risk by Avseth et al. (Cambridge, 2005). Relationships
between compressional-wave and shear-wave velocities in clastic silicate rocks by Castagna et al. (Geophysics, 1985). Shale
rock physics and implications for AVO analysisA North Sea
January 2009

The Leading Edge

107

Downloaded 04/15/15 to 72.37.140.34. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

R o c k

p h y s i c s

demonstration by Avseth et al. (TLE, 2008). Quartz cementation in sedimentary basins by Bjorlykke and Egeberg (AAPG
Bulletin, 1993). Well-log and seismic data analysis using rock
physics templates by degaard and Avseth (First Break, 2004).
Quartz cementation in oil eld sandstones: a review of the key
controversies by Worden and Morad (in Quartz Cementation
in Sandstones, 2000). The origin of large-scale quartz cementation: Evidence from large data sets and coupled heat-uid mass
transport modeling by Giles et al. (in Quartz Cementation
in Sandstones). Modeling quartz cementation and porosity in
reservoir sandstones: examples from the Norwegian continental
shelf by Walderhaug et al. (in Quartz Cementation in Sandstones). Petrophysical and petrographical analysis of quartz cement volumes across oil-water contacts in the Magnus Field,
northern North Sea by Barclay and Worden (in Quartz Cementation in Sandstones). Quantifying information and uncertainty
of rock property estimation from seismic data by Takahashi (doctoral dissertation, Stanford, 2000). Estimation of net-to-gross
from P and S impedance in deepwater turbidites by Vernik et
al. (TLE, 2002). Detection of gas in sandstone reservoirs using AVO-analysis: A 3D seismic case history using the Geostack

108

The Leading Edge

January 2009

technique by Fatti et al. (Geophysics, 1994). AVO attribute


inversion for nely layered reservoirs by Stovas et al. (Geophysics, 2006). Poroelastic Backus averaging for anisotropic layered
uid- and gas-saturated sediments by Gelinsky and Shapiro
(Geophysics, 1997). Statistical uncertainty of seismic net pay
estimations by Connolly and Kemper (TLE, 2007). The Rock
Physics Handbook by Mavko et al. (Cambridge, 1998). Mapping lithofacies and pore-uid probabilities in a North Sea reservoir: Seismic inversions and statistical rock physics by Mukerji
et al. (Geophysics, 2001). Combined porosity, saturation and
net-to-gross estimation from rock physics templates by Avseth
et al. (SEG 2006 Expanded Abstracts).
Acknowledgments: Thanks to Lundin-Norway for providing data
used in this study. We thank Hans Oddvar Augedal at LundinNorway for valuable input on the geology and diagenesis of the
studied case. Also thanks to Anders Drge at StatoilHydro for
valuable feedback.
Corresponding author: per.avseth@odin-petroleum.no

You might also like