utah, bartlesville, company parkway...would work well if the rock consisted of only two components,...
TRANSCRIPT
USING A MULTIPLE LOG APPROACH TO EVALUATE GREEN
RIVER OIL SHALE IN THE PICEANCE CREEK BASIN
Rob Habiger
Phillips Petroleum Company157 GB
Bartlesville, OK 74004
R.H. Robinson
Phillips Petroleum Company8055 E. Tufts Ave. Parkway
Denver, CO 80237
ABSTRACT
A method has been devised for predicting modi
fied Fischer assay yield from conventional loqs in
the Piceance Creek Basin of Colorado. The tradi
tional approach for using well logs to predict oil
shale richness has been to relate one of the poro
sity tool responses to the modified Fischer assay.
This approach is unsatisfactory when there is sig
nificant lithology variation in the oil shale rock.
This study involves an exploration area of 86 Km?
in Rio Blanco County, Colorado. Fifteen coreholes
were drilled on the property to provide the data
base for development of the method. The method
describes the rock in terms of a simple litholoqy
model consisting of: 1) carbonate, 2) clay, 3)
porosity, and 4) organic matter. Since each of
these quantities can vary independently, the simple
approach of using only one porosity tool does not
work. Instead, the density, sonic, and resistivity
logs are used in combination to isolate the amount
of organic matter by predicting the modified
Fischer assay. Problems related to log data vari
ability and core data quality will also be dis
cussed.
INTRODUCTION
A suite of conventional wireline logs is used
to predict oil shale richness on a property of 86
Km2 in Rio Blanco County, Colorado. The project
location is shown in Figure 1. Althouah there
exists a variety of correlation equations in the
literature relating well log measurements to oil
yields in Colorado and Utah, none are satisfactory
in predicting yields in this property. The assay
prediction method we developed uses data from 15
wells that were both cored and logged. The method
was applied to predict yield in 7 wells that were
only logged, and were previously unusable for
resource evaluation and mine plannina. For one of
these holes the log data was too poor to use.
An additional result of this work is a regres
sion equation that predicts true bulk density of
the rock from logged density. This relationship is
not straightforward since well logs give an indi
rect measurement of bulk density and overestimate
it in organic-rich material. In addition, logged
density exhibits greater variability than the core
measured density. These factors require that the
log data be carefully adjusted and correlated with
the core data to provide an accurate bulk density
for use in mining calculations.
REVIEW OF PRIOR WORK
Table 1 lists most of the published work con
cerning use of well logs to predict oil shale yield
or more specifically, the Modified Fischer Assay
(MFA) in Colorado and Utah. In each case, the data
base for these correlations is quite limited. Fur
ther, the correlations are based upon a sinqle
variable, usually the logged bulk density or forma
tion velocity. Another point, more subtle but nev
ertheless very important, is that some correlations
were developed from log data while others were
developed from core data.
In the past, any of three porosity logs (den
sity, sonic, and neutron) were used independently
to predict oil shale yield. Any of these loos
45
MAHOGANY SHALE
PROJECT>Nfc**e
*/
COLORADO
FIG. 1
Locations for literature correlations (see Table 1)
reference (3) is in the Piceance Creek Basin (exact location unknown)
IDAHO
UTAH
WYOMING
Green River Basin
COLORADO
ICreek
FIG. 2
46
TABLE 1
Literature Correlations for Predicting Oil Yieldin Colorado and Utah Oil Shales
REFERENCE EQUATIONA
(1) Smith (1956):
(a) Y =
(31.563)(p)2-(205.998)(p)+(326.624)(b) Y =
(22.881)(p)2-(166.8)(p)+(280.439)
(2) Bardsley and Algermissen (1963):
Y =
Y
= (-66.41)(PH(171.23)= (41.01xlO-4(At)2-(16.71)
(3) Tixier and Alger (1967):
Y = (-59.43)(p)+154.81
(4) Cleveland-Cliffs (1975):
Y = (496.325)(p)-^.(285.176)Y = (l56.767)(At)l-8.(29.1703)
REFERENCE LOCATION6
(1) (a) Garfield County, Colorado
(b) NE corner of MSP property
(2). .Unita Basin, Utah
(3) Piceance Basin, Colorado
(4) Unita Basin, Utah
REFERENCE DATA TYPE DATA QUANTITY YIELD RANGE
(1) Lab Grain Density 1 corehole 3-78 GPT
(2) Schlumberger Log 1 corehole 5-80 GPTC
(3) Schlumberger Log 1 corehole 2-37 GPT
(4) Birdwell Log 3 coreholes 0-80 GPT
A. Units are: p in gm/cc, At in microsec/ft, Y in gal/ton.
B. See Figure 2 for relative locationsC. Data were not fit well at the higher yields (underpredicted).
TABLE 2
Summary of Log and Assay Data
DRILL HOLE LOG DATA
BPB SCHLUMBERGER PHILLIPS R&D
-MFA
P-1 X X
P-2 X
P-3
P-4
P-5 .X
P-6 .X x X X
P-7 .X
P-8 .x
P-9 .x X
P-10 .x X X X
Sungas FC-36B-1-100.
P-12 .X X
P-13 .X X
P-14 .X X
P-15 .X X
P-16 .X X
P-19 .X X
P-20 .X X
P-22 .X X
P-27 .X X
P-28 X
Sungas 24-1-100 X
A. The log data quality was too poor for use.
47
would work well if the rock consisted of only two
components, a host matrix and orqanic matter. The
organic matter volume fraction would represent
porosity to these logs. A single variable correla
tion could then be made between MFA and loqged
porosity. Althouqh this single variable approach
may work for limited data, it is inadequate for our
project area. In Figures 3, 4, and 5, we plot MFA
against each porosity loq. The data for these
plots come from coreholes spanning the entire prop
erty. These plots are a 4% sample taken from all
the log and core data. Our correlations include
the entire data base which is described in the next
section. In the Figures, note the considerable
point scatter due to a variety of factors. Miner-
alogic variations are probably chief amonq these
factors, but other factors such as hole condition
and log calibration are also important. Another
scatter plot, Figure 6 shows MFA plotted against
logarithm of resistivity. There is evidence for a
relationship but point scatter is again very large.
It is quite clear that none of these logs, by them
selves, are sufficient to predict MFA with the
desired accuracy. In a later section we will show
how log parameters can be combined to predict MFA
quite accurately.
Density is the most important parameter for
predicting oil shale richness, but one must ensure
that there is no ambiguity in its measurement.
There are three different densities reported in the
literature in connection with predicting yields
from oil shale. The first of these densities is
grain density which is measured from ground rock
samples (Smith, 1956). Similar work has been done
at the Phillips Research Center (Kokesh, 1982).
The results of the Phillips work are shown in Fiq-
ure 7 along with Smith's correlation. The two
pieces of work agree quite well. However, this
grain density cannot be applied directly to a mea
surement made from a borehole. The borehole mea
surement more closely approximates a bulk density
which in turn is related to the grain density by
equation (1) below.
Pb= U -4>) P9 + ?Pf f1)
where
<}> = fractional porosity
pd= bulk densiy of whole rock
pg= grain density of ground rock
Pf= density of the fluid in the
pore space
Bulk density is the density of the total rock
which includes any pore or fracture space. This
space is usually filled with a fluid such as water,
oil, or gas. When the rock is ground, any fluid
escapes and density of the ground rock is grain
density. The difference between these two den
sities can be quite significant as seen in the sec
ond track of Figure 8. These densities were used
to calculate porosity in the third track using
equation (1). Incidently, notice that the resis
tivity indirectly correlates with the porosity.
This behavior is consistent with the model that
porosity is salt water filled. In addition, the
confusion regarding density is further complicated
by the density logging tool measuring an"apparent"
bulk density. This apparent density can varysiq-
nificantly from the true bulk density in orqanic
rich material (Habiger, 1981). This topic will be
discussed in detail in a latter section of this
paper.
DATA BASE
The data for this study can be separated into
three sets. One set comes from our prospect's 1980
exploration season, a second set from the project's
48
80 r
60
co
.2 40>-
20
0
1.8
*x x
$ x x x xxxx * * * *
X X j(X
Xx
* xXx<x xxx,
xxiv x* xx XJC xr
* ic * * x x. X
-x xv
xx x
- -
X T^XXx * xX MX Xjc X X
*
XJ* * * x
XJ(xx
**^ X*X*X
x
"x x
J** x "x*** *
*
x
xx'xx* v 3c x& x *"*
.>
xx
Tv ^J"* "xw1 , x x X^ ^.x^ *X| g ^x/,
2.0 2.2 2.4
Density (gm/cc)
FIG. 3
2.6 2.8
80r
c
o
T 60
CO
a>
?40
20
130
* X.
K *X
X X
X x
*
x*.
**K
# X*** "-*
x T"
x_
^ # * XX
x x****
X
^ xv -
XxX
Xx
110 90 70
1/v (microsec/ft)
FIG. 4
50 30
49
80 r
c
o
CO
60
2
.2 40
20
X
X
X
X X
X
XX
**
*X
x*t x
x xxXX X
*
X
X>oc x
X X X
X
xxx
xX X
"
x x *< x
x>cx
x X x
x #x
xxxx
x Xx Xx
*
X X X X
**"
XX
V
^X
x x *xxx
* **.X
x X xx
x
x*
x
x*X
Xx x Xx
xx
xX
xx
X
* x x
xx Xx X
*X *
I 1 Xj
xx *xx
_.xx xi xx, x, X 1
40 80 120
Neutron (standard neutron units)
FIG. 5
160 200
80
o4 60">
CO
T3
0)
>-40
20
oL
5
x5 xr
.XXXX X \ jf
*"
X *x.
xx \
x *
xXX..* x#Xxx* x
XX ""i'
x xxxx
xx
~
xxx
X"Xxxl " ^l* *x"<A
"X
x x* J" * X
X * XXx x o,
* X X*w
XV%X*^
xxxx x x
xJ<
xx5c .,
4 3 2
Log(R) R in ohm meters
FIG. 6
50
80 rSmith's equation (b)
c
o
CO
O)
35
0)
1.8 2.0 2.2 2.4
Grain Density (gm/cc)
FIG. 7
2.6 2.8
FIG. 8
51
1981 exploration season, and the third set from two
Sungas drill holes that were loqqed by Phillips.
The logging service company was the British com
pany, BPB Instruments, Inc. In addition, Phillips
Research Center personnel recorded full sonic wave
forms in three holes. Our full waveform sonic log-
ging technique requires that Schlumberger Well Ser
vices equipment be on location concurrent with the
Phillips operation. While there, Schlumberger ran
a limited log suite for those three holes. This
additional logging provided a useful check on qual
ity and calibration of the BPB loqs. All core was
analyzed by the MFA and there were some special
analyses performed on selected core.
ADJUSTMENTS TO LOG DATA
Averaging and Caliper Log Analysis
Core samples were assayed on 1-foot intervals
in the 1980 season and on 2-foot intervals in the
1981 season. In addition, log data were recorded
digitally every 6 inches. For the purpose of mak
ing correlations, a filter function averaged the
log data so that it would match the vertical reso
lution of the core data.
The caliper log was used to eliminate those
parts of the drill holes too rough for an accurate
log density measurement. Unfortunately, density
log is both the most important log and the one most
affected by borehole rugosity. The density tool
must be held against the borehole wall during log
ging. Borehole irregularities cause gaps between
the tool and the borehole wall which leads to erro
neously low density values. The caliper log
locates these irregularities and the predicted
yields from those depths are not used in the corre
lation.
Log Calibration and Normalization
Ideally, all logging tools would be properly
calibrated and repeatable. However, Figure 9 shows
a series of density log histoqrams that exhibit
shifts of as much as 0.15 gm/cc relative to each
other. Lange (1980) reported that the density log
is sometimes miscalibrated and this could explain
the observed shifts. In addition, there could be
other reasons for the shifts, such as hole-to-hole
variations in rock type. Table 3 shows the results
of an analysis of variance for the various logs and
yields among the 1981 core holes. All the logs and
yields have statistically significant variability
from hole-to-hole, but the variability of the den
sity log is by far the largest. Without studying
the rock and practices of the logging company in
detail, we cannot explain the source of this large
variability. Nevertheless it is critical to cor
rect for this large density variability since den
sity greatly affects predicted yield. After trying
many correction methods, we settled on the simple
approach of transforming to a new variable.
Density Variation = (log density)-(loq densitymean) (3)
In this equation, the log density mean is calcu
lated for each hole and listed in Table 4. Unfor
tunately this transformation removes all the den
sity variability, even that which is due to average
yield variation. It is acceptable though because
the grade variability is a small part of total
variability.
DATA INTERPRETATION AND ANALYSIS
Geologic Model
We looked for some natural divisions of the
data which would reduce the variability between MFA
52
X OF
SAMPLE
5.00 +
2.50
0.0
5.00
2.50
0.0
2.50
0.0
5.00
2.50
0.0
5.00
2.50
0 +...
1.80
Log Density Histograms (gm/cc)
P-28 **
*****
*******
* * * ** ************
** a*****************************************
P-10* * * *
** * **
P-3
****
** * *** ***
* ** *****************************************
P-14
* * * * *
* * ** ** * *
* * * *
********* ******** * **
* ******************************** **********
* *** ********************************************* v?*
P-27
** * * *
* * ***************************************
1.90 2.00 2.10 2.20 2.30 2.40 2.50 2.6
FIG. 9
53
TABLE 3
Analysis of Variance procedure on Log and Yield Data
Data Base: 1981 Upper Zones
Model: log density = well #
Source DF Mean Square
Model 4
Error 901
0.740
0.0095
F Value
77.93
PR>F
0.0001
Model: log A t = well #
Model 4 546.976
Error 901 133.410
4.10 0.0027
Model: logarithm (log resistivity) =well #
Model 4 3.914 9.62
Error 901 0.407
0.0001
Model: Yield = well #
Model 4 245.330
Error 901 93.082
2.64 0.0329
Data Base: 1981 Upper and Lower Zones
Model: log density = well #
Model 9 0.688
Error 2076 0.010
Model: Yield = well #
Model
Error
9
2076
303.998
87.618
66.80
3.47
TABLE 4
Mean Logged Densities by Core Hole
0.0001
0.0003
Hole
P-12
P-13
P-14
P-15
P-16
P-19
P-20
P-22
P-27
P-28
Mean Density
2.32 gm/cc
2.19
2.35
2.30
2.28
2.22
2.25
2.21
2.23
2.20
54
and well logs. A useful, yet simple, division sep
arates the oil shale strata into an upper and a
lower zone. This division is easily picked from
logs as the example in Figure 10 shows. Even
though the densities in the two zones are generally
the same, the lower zone has lower resistivities
and slower sonic velocities. Although mineraloqy
of the western oil shale is probably quite compli
cated, a relatively simple model can be used to
interpret the log data and arrive at an oil yield
prediction. The vertical division that is so
apparent from the logs provides the key to this
model. We will refer to this division throughout
this report by an upper and lower zone.
The mineralogic model consists of four compo
nents: 1) carbonates, 2) clay with associated
bound water, 3) porosity, and 4) orqanic matter.
Figure 11 shows neutron-density crossplots for the
upper and lower zones of P-10. In the upper zone,
the points fall between limestone and dolomite
matrix lines. In the lower zone, the points shift
in the direction indicating addition of clay to the
matrix. This interpretation means the lower zone
has considerably more clay than the upper zone and
is supported by a number of observations:
1) A basin-wide geologic feature, the Blue
Marker, separates the strata into the Parachute
Creek Member and the Garden Gulch Member of the
Green River Formation (Newman, 1980). The two mem
bers are distinguished by a visible change in clay
content. X-ray diffraction and thin section data
for hole P-6 are listed in Table 5 and support the
model of a carbonate-clay matrix that increases in
clay content in the lower zone (Basick, 1982).
2) Figure 12 compares histograms of the amount
of water collected during the Fischer Assay for
corehole P-15 in upper and lower zones. The
increased amount of water in the lower zone is
probably due to the presence of bound water in the
clays. This increase explains why the lower zone
is characterized by a lower resistivity.
3) It is also reasonable to expect that the
clay causes the compressional velocity to decrease
in the lower zone. This effect occurs as shown on
the sonic log in Figure 10.
4) Additional evidence for increased clay con
tent in the lower zone is provided by full waveform
sonic data taken by Phillips R&D personnel (Basick,
1982). Figure 13 shows the full waveform data for
hole P-6. Two characteristics stand out in the
waveforms. The first onset, which is the arrival
of the compressional wave, shows decreased velocity
in the lower zone which was apparent in Fiqure 10.
In addition, the shear wave disappears in the lower
zone. The clay index (to be discussed later) cor
relates with both effects. Clay decreases the com
petency of the rock resulting in slower compres
sional and shear velocities. In the clay-rich
zones, shear velocity most likely becomes less than
borehole fluid velocity which results in no
recorded shear wave. This effect most probably
causes the disappearance of the shear wave. In
addition, both compressional and shear waves will
be more highly attenuated in the softer rock. The
interval at 795 is uncharacteristic of the lower
zone in that it has a low clay content as evidenced
by the clay index, the compressional onset time,
and the strong shear wave amplitude.
Clay contains water that is chemically bound
in the rock. But there is also a considerable
amount of water that is "free", i.e. contained in
the rock pore space. The work of Kokesh (1982),
which compared bulk and grain densities, showed
clearly the existence of porosity. This porosity
55
P-15
FIG. 10
TABLE 5
X-Ray Diffraction Data for P-6
Depth Quartz
25
Peak
Analcime
4
He ights
Ankerite
Dolomite
30
Calcite
80
Area
Illite
Mixed Layer
2,8A
Pyrite
7
Peak
Albite
Heights
Muscovite
0
X
11
Y
326 4 0
342 35 41 34 3 6 5 6 0 5 0
668 23 14 45 19 19 0 3 0 0 0
742 32 9 9 20 24 14 2 10 18 0
753 7 0 79 0 1 0 3 0 6 1
807 43 0 21 0 5 0 3 0 16 11
1019 85 0 4 16 24 0 9 10 11 8
1106 24 11 55 0 40 10 5 0 0 0
Numbers s
suspected
hould only be compared within a
to be K-Feldspar. Samples at
mineral type. Minerals X and Y are not identified although
^53 and 807 are not typical of the interval.
X is
Montmorillonite
56
P-10 Schlumberger log data
*0 6o
Neutron Porosity (%)
(a)
upper zone
o 22
/ ".lower zone
/ " ****
"
.
/ m f* mJZ^n**
/ ** AS**"
'""%/** * #c' >v
V*
/ "/",
V
V6./
I i i i i i i i i
Neutron Porosity (%)
(b)
FIG. 11
57
P-15
upper zone
X OF
SAMPLE
+ + + + + ? + ? ? ?? 0.0
B.BB .50 1.00 1.50 2.00 Z.50 3.00 3.50 4.00 4.50 5.00
.50 1.00 1.50 Z.00 2.50 3.00 3.50
Water from Fischer Assay (wt %)
4.00 4.50 5.
FIG. 12
58
Full Waveform Sonic
ml
50 0
upper zona
lower zone
700
800
.'h
"IMffifi*.:.:.Zk r.i, . 1; -;r. n r*
mmm
P-6
FIG. 13
59
cannot be distinguished from organic matter by the
standard porosity logs (density, sonic, neutron).
However, the porous intervals generally have low
resistivities which suggest that they are filled
with conducting water.
Log Interpretation
The density log would probably be a reliable
indicator of organic matter content if it were not
for the porosity variations. Grain density mea
surements indicate that the clay and carbonate have
approximately the same density so that the volume
fraction of organic matter (Vo) could be approxi
mated from the bulk density (pb) and matrix density
(pg).
Vo = (pg-pb)(pg-D"
(4)
In this case, the rock-matrix density is fairly
constant no matter what the mixture of carbonate
and clay. However, porosity causes density to vary
considerably and the simple relation in equation
(4) is not valid. The resistivity loq can be used
to identify the porous intervals. Unfortunately,
the resistivity log cannot tell the difference
between porous intervals filled with free water and
clay rich intervals.
This inability to distinguish porosity and
clay on the resistivity log can be demonstrated by
a series of crossplots. Figure 14 compares cross-
plots of yield versus density variation with dif
ferent resistivities for a composite of the upper
zone. Notice that the hiqhest resistivity inter
vals show a good yield-density correlation. These
intervals have very little clay or porosity. As
the resistivity decreases, points scatter towards
lower densities. This is due to increased poros
ity. Still there are some points that remain on
the correlation line even at the lower resistiv
ities. These points owe their low resistivities to
clay and not porosity.
The clay-porosity ambiguity can be resolved if
the clay content can be determined independently.
Probably the most common clay indicator used in log
analysis is natural gamma ray activity. But for
this project, the qamma ray curve has much charac
ter but does not appear to be related to clay con
tent. Therefore it is necessary to use other logs
to determine clay content, and we have already men
tioned that the sonic and neutron logs are affected
by clay. Unfortunately, the amount of neutron data
on hand are very limited since no neutron logqing
was done in 1981 and the 1980 neutron log data are
incomplete. Consequently, we chose the sonic log
for clay volume determination. Figure 15 displays
sonic-density crossplots for two different resis
tivity ranges while the higher resistivity range
excludes intervals with significant clay content,
the lower resistivity range includes intervals with
either clay, porosity, or both. These data are
taken from a composite of the upper zones. The
correlation line established from the clay-free
graph is drawn on the low resistivity graph. Those
points that still fall on the correlation line are
interpreted as clay-free, porous intervals. On the
other hand, points that fall to the lower right of
the line are intervals with varying amounts of
clay. Using the sonic-density crossplot, we assign
a clay index to each point based upon its departure
from the clay-free line.
CI = (log At) + (127.31)DV -84.84 (5)
We found this index to be most useful in the upper
zone to distinguish porosity from clay in a predom
inately carbonate matrix.
60
100 r
80
~ 60
> 40
20 -
high resistivities
4 < log R ^ 5
"\
-
-
- ^^C
i i i ^S. _i 1 1
-0.5 0.3 -0.1 0.1
Density Variation (gm/cc)
0.3 0.5
100
-0.5
low resistivities
0 ^ log R S 2.5
0.3 -0.1 0.1
Density Variation (gm/cc)
FIG. 14
0.3 0.5
61
40 r
60high resistivities
80
100
120
140
-0.5 -0.3 -0.1 0.1 0.3
Density Variation gm/cc)
0.5
40
60
low resistivities
80
Ew 100
120
140
y s*Vl&" cav
*"""""
" 1
-0.5 -0.3 -0.1 0.1 0.3
Density Variation (gm/cc)
FIG. 15
_i i
0.5
62
Our interpretation is supported by showing in
Figure 16 that the points of Figure 14 with the
highest clay indices fall on the correlation line
while the points with the lowest clay indicies
(porous intervals) fall off the correlation line.
Clay content by itself does not affect the
density-to-yield correlation. It is porosity that
has the most adverse affect on the correlation.
However, porosity can only be detected from logs by
measuring formation resistivity, and resistivity is
strongly affected by clay. Therefore, the presence
of porosity requires that clay content be quanti
fied so that resistivity can then be effectively
used to correct for porosity on the density log.
In effect, the clay index describes how much of the
resistivity decrease is due to clay and how much is
due to porosity.
By using density, sonic, and resistivity logs,
we have the information necessary to identify indi
vidual contributions from each of four main compo
nents: 1) carbonate, 2) clay, 3) porosity, and 4)
organic matter. The remaining task is to find a
suitable mathematical model for expressing resis
tivity as a function of these rock components. One
approach would be to calculate porosity as true
porosity + organic-matter volume fraction from the
density log. Then this porosity and clay volume
fraction could be applied to a standard loq analy
sis used for oil & gas applications. Next an "oil
saturation"
would be calculated which could be
related to the MFA assay. However, this method
requires estimation of five parameters: two forma
tion factor constants, a saturation exponent, an Rw
and a parameter relating clay volume to clay index.
Since this method is not a standard application of
accepted oil & gas techniques, we were hesitant to
assume that common values for these parameters are
acceptable. Consequently, we feel that it is much
more scientifically sound to simply do a multiple
linear regression using density variation, loga
rithm of resistivity, and clay index. This three
variable regression is used for the upper zone. In
the lower zone, good prediction was obtained using
only two variables, density variation and logarithm
of resistivity. The results of these correlations
using the 1981 core data are:
Upper Zone
Y = (-74.37)(DV)+(7.86)(log R)+(0.5)(CI)-9.65 (6)
Lower Zone
Y = (-81.58)(DV)+(4.70)(log R)+9.36 (7)
These equations predict yield in gallons/ton when
the variables are calculated as described in equa
tions (3) and (5) with the log measurements of den
sity, At, and resistivity in gm/cc, microsec/ft,
and ohm-m. Figure 17 graphs core MFA versus log-
predicted yields using equations (6) and (7) for
the 1981 data. Points eliminated from the regres
sions come from three sources: intervals where the
caliper indicates an unreliable density log, one
large interval in P-16 where MFA do not depth-
correlate with any logs, and short intervals where
MFA are insufficient to be representative of log
measurements.
PREDICTING TRUE BULK DENSITY
The density log indirectly measures true for
mation bulk density. Actually this loq value is an
apparent density that is generally hiqher than the
true bulk density in organic rich rock. This sec
tion describes the relationship between logged den
sity and density measured on core samples. Because
our depths are shallow, we assume that laboratory
values represent the in situ density very closely.
Extensive laboratory bulk density measurements were
63
uu
80
low resistivities
high clay indicies
60
"\
\
40
X
20
0 ' , ,
*
.
" 7\T-0.5 -0.3 -0.1 0.1 0.3
Density Variation (gm/cc)
0.5
100
low resistivities
80 low clay indicies
60
a
a'
2 ^x
>40
20
,^
0 i i Vb
# ^n.
-0.5 0.3 -0.1 0.1
Density Variation (gm/cc)
0.3 0.5
FIG. 16
64
20 40 60
Log Predicted Yield (gpt)
100
FIG. 17
2.8 r
P-15
2.6
2.4 -
2.2
2.0
xxx
&yXX Mx**<
**Jr*
NsxJRx $X f X J&***m
xvx **jyiJFxxfSS
x x^Sxx%JBJkx
)t apocjtiry,..^ l^Fxx
xx5c SiAxx i"xfMxX
>x*?)W~X
X3*v*tt<WS*x*xx><
XX / X X< XX X M
3KX.3O0OCXX X
X Tf/T^PBysc $
/xA x$ xxx
_yx J*b X XX
s ******* xx*"
* r $x*Vx \*$XX XXM
xX
I
XjSX x*
x"x? xx*xm
x** x
x*xxx
X
5xx sX XXX X V,
xwx)T KX*
X
1.8
1.8 2.0 2.2 2.4
Log Density (gm/cc)
FIG. 18
65
2.6 2.8
made on two foot core intervals from holes P-15,
P-22, P-27, and P-28.
An example of the relationship between loq
density and bulk density is shown in Fioure 18.
The relationship between bulk density and loo den
sity is further complicated by the variability in
log densities that was discussed earlier. An anal
ysis of variance for the log and lab densities is
listed in Table 6 and show that log data hole-
to-hole variability is greater than lab density
variability even though the lab density variability
is significant. Figure 19 plots log density versus
lab density for the entire data base. The log den
sity variability scatters the points so that the
departure from the linear trend at low density is
less apparent than it is in Figure 18. The corre
lation for predicting true bulk density increases
by using the same density variable qiven in equa
tion (3). Fiqure 20 shows the results for thistransformation. The prediction equation is
bulk density = (2.227) + (1.189)(DV) (8)
TABLE 6
Analysis of Variance Procedure for Log and Lab Density
Data Base: Coreholes P-15, P-22, P-27, P-28
Model: Log density = well #
grateful to Phillips Petroleum Company for permis
sion to publish these results.
Source
Model
Error
DF Mean Square F Value PR>F
3 1.357
1718 0.011
121.72 0.0001
Model: Lab density = well #
Model 3 0.358
Error 1718 0.015
23.28 0.0001
ACKNOWLEDGEMENTS
We acknowledge the assistance of J. T. Basick
in providing full waveform sonic data and analysis
of the x-ray diffraction data. Hans Schmoldt pro
vided much supporting technical material. We are
REFERENCES
Bardsley, S. R. and Algermissen, S. T., 1963,
Evaluating oil shale by log analysis: Journal
of Petroleum Technology, vol. 15, no. 1, p. 81-
84.
Basick, J. T., 1982, Personal commun.
Berge Exploration, Inc., 1980, Mahogany Shale
Project: Report for 1980 Field Season (Unpub
lished Report).
Cleveland-Cliffs Iron Company, 1975, Geophysical
logging report for the White River Shale
Project (Unpublished Report).
Habiger, Robert M., 1981, Phillips Petroleum
Company Research and Development Report (Unpub
lished).
Kokesh, Fritz, 1982, Personal commun.
Lange, W. H., Jr., 1980, SPWLA ad hoc log calibra
tion committee report: Porosity log calibra
tions: The Log Analyst, vol. 21, no. 2, March-
April, p. 14-19.
Newman, K. R., 1980, Geology of oil shale in the
Piceance Creek Basin, Colorado: Rocky Mountain
Association of Geologists Symposium, p. 199-203.
Smith, John Ward, 1956, Specific gravity-oil yield
relationships of two Colorado oil shale cores:
Industrial and Engineering Chemistry, vol. 48,
no. 3, p. 441-444.
Tixier, M. P., and Alger, R. P., 1967, Log evalua
tion of nonmetallic mineral deposits: SPWLA
Symposium Paper R.
66
2.6
E 2.4
2.2
2.0
P-22
P-27
P-28
1.8
1.8
'xlLxSx
<x#*
*%xxxxKxx3*xV*Sc
x
x x S8 y-JTxjC***"*
*v*
/*Jx X*
/
2.0 2.2 2.4
Log Density (gm/cc)
FIG. 19
2.6 2.8
2.8 r
-0.3 -0.1 0.1
Density Variation (gm/cc)
FIG. 20
67
0.5