Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

PETROLEUM TRANSACTIONS

~ f l f i
1<'.1'6'.11 t"I>"
A Statistical Reservoir-Zonation Technique
J. D. TESTERMAN
ABSTRACT
A statistical technique to identify and describe naturally
occurring zones in a reservoir and to correlate these zones
from well to well is described. The technique is partic-
ularly useful in describing a reservoir where crossfiow
between adjacent strata is important in determining reser-
voir behavior. Although it has been used primarily for
permeability zonation, the technique is general and can
be used to correlate any reservoir property or related data,
such as the information contained in well logs.
INTRODUCTION
One of the first problems encountered by the reservoir
engineer in predicting or interpreting fluid displacement
behavior during secondary recovery processes is that of
organizing and using the large amount of data available
from core analysis. Permeabilities pose particular prob-
lems in organization because they usually vary by more
than an order of magnitude between different strata. Due
to the sheer volume, it is almost always necessary to group
data and to use an average value to represent a number
of measurements. Perhaps the most common method now
used to group permeability data is the capacity-fraction
technique, which ranks permeabilities in order of magni-
tude, regardless of the physical location of the permeabili-
ties within the reservoir. The cumulative per cent capacity
is plotted against cumulative per cent thickness. This plot
is divided into an arbitrary number of zones, generally of
equal thickness. Five zones (or averaged groups of data)
usually are obtained, each of which is treated as homo-
geneous in subsequent calculations. The division so ob-
tained has no physical meaning; strata in the same zone,
calculation-wise, are usually not adjacent in the reservoir.
Reservoir engineering techniques being developed will
handle crossflow that occurs between adjacent communi-
cating reservoir strata because of imbibition and gravity
segregation. Since crossflow occurs between physically
adjacent layers within the reservoir, a new zonation tech-
nique recognizing the actual location of strata within the
reservoir is necessary. Similarly, the recognition of natural
zones is important for predictions of oil recovery by proc-
esses involving diffusion. One such process is miscible
displacement, where predictions of lateral diffusion within
the reservoir must recognize the actual location of the
Original manuscript received in Society of Petroleum Engineers office
Feb. 28, 1962. Revised manuscript received June 29, 1962. Paper pre-
sented at SPE Production Research Symposium, April 12-13, 1962, in
Tulsa, Okla.
Discussion of this and all following technical papers is invited. Dis-
cussion in writing (three copies) may be sent to the office of the
Journal of Petroleum Technology. Any discussion offered after Dec. 31,
1962, should be in the form of a new paper. No discussion should ex-
ceed 10 per cent of the manuscript being discussed.
AUGUST, 1962
SPE 286
JERSEY PRODUCTION RESEARCH CO.
TULSA, OKLA.
invaded zones in relation to the rest of the formation.
Natural zones must also be adequately recognized to
account for heat transfer within the reservoir during the
thermal exploitation.
Because of the complexity of the problem, statistics
appear to offer the only practical hope of dividing a reser-
voir into physically-meaningful natural zones. This paper
presents a statistical technique for identifying these natural
zones and for ascertaining which ones are likely to be
continuous between adjacent wells. The zones defined have
minimum variation of permeability internally and a maxi-
mum variation between zones. The technique is general
and can thus be applied to reservoir properties other than
permeability.
The method will guide the reservoir engineer in esti-
mating which zones are likely to be continuous between
wells. However, a statistical correlation based on perme-
abilities in two different wells is no guarantee that the
zones so defined are, in fact, continuous. Rather, the as-
sumption of continuity must be consistent with geological
data concerning the depositional environment, as well as
justified on the basis of engineering judgment in combina-
tion with statistics, just as judgment is required with
conventional zonation methods.
CALCULATION PROCEDURE
The reservoir zonation technique is a two-step opera-
tion. The steps are individually described, and a sample
calculation is presented in the Appendix.
ZONATION OF INDIVIDUAL WELLS
First, the set of permeability data at a single well is
divided into zones. These zones are selected so that varia-
tion is minimized within the zones and maximized be-
tween the zones. The equations'" used to zone the data are
1 [L __ ]
B = L _ 1 i:: 1 nJ, (k,. - k. Y , .
(1)
1 [L "<I _]
W = -_- ~ ~ (k;j - k
i
,)' ,
N L i=lj=l
(2)
and
B-W
R=---,
B
(3)
where B = the variance between zones, L = the number
of zones, i = the summation index for the number of
zones, j = the summation index for the number of data
within the zone, m
i
= the number of data in the ith zone,
r:. = the mean of the permeability data in the ith zone,
k.. = the over-all mean of the data in the well, W = the
pooled variance within zones, N = the total number of
4References given at end of paper.
889
data, the k,/s = the permeability data and R = the zona-
tion index.
The first step in the calculation is to divide the perme-
ability data, in their original order of depth, into all pos-
sible combinations of two zones. Eqs. 1 and 2, and then
Eq. 3, are calculated for each of these possible two-zone
combinations. Eq. 3, the index of zonation, is the criterion
used to denote the best division. This index, which ranges
between 0 and 1.0, indicates how closely the division cor-
responds to homogeneous zones. The closer the index is
to 1.0, the more homogeneous the zones. Therefore, the
larger index denotes the best division into two zones and
is retained for comparison with other indices.
After the best two-zone combination is determined, the
data are divided into all possible three-zone combinations,
with the previous two-zone point of division being one
of the (two) points of division into three zones. Eqs. 1,
2 and 3 are again computed for each of these divisions,
and the best division is again defined by the maximum
index.
Next, the data are divided into all possible four-zone
combinations. The previous points of division are again
retained as two of the (three) points of division into four
zones. In this manner, the set of permeability data can
be divided into any number of zones up to the number
of data.
At each extension of the number of zones in a well,
the new index is compared with the previous index. The
division into additional zones continues until two succes-
sive indices show no significant increases. Our work with
reservoir permeability data has indicated that the differ-
ence between two successive indices is not significant if
it is less than 0.06.
To demonstrate the zonation technique, we selected
permeability data from two wells. These data and the
zonation results are presented in Table 1. Part (a) shows
the indices for all possible two-zone combinations. Indices
for the possible three-zone and four-zone combinations
are given in Parts (b) and (c), respectively. Table 2
gives the final zonation of the data and the corresponding
average permeability for each well zone. In Part (a) of
Table 1, for example, the best two-zone index of the five
possible in Well No. 1 is 0.99. Because this index is so
close to 1.0, further partitioning is pointless. The well is
best described as two zones. This is substantiated by cal-
culating the three-zone index of 0.99. Well No. 2 has
0.92 for its best two-zone index. The best three-zone
index of 0.99 shows that the well is better described as
three zones than as two. Again, the 0.99 index indicates
that no further partitioning is necessary. The indices of
0.99 obtained for all possible four-zone divisions sub-
stantiate the three-zone description and show there is no
reason to subdivide beyond three zones.
CORRELATION OF ZONES
BETWEEN ADJACENT WELLS
After individual wells are zoned, the second portion of
the calculation is undertaken. This part correlates the
zones from well to well throughout the reservoir to aid
the engineer in determining continuity of the strata. The
correlation is based on a statistical comparison of the
difference of means of two zones in adjoining wells with
the difference that could be expected from variation of
measurements within the zones. If the difference of means
is less than or equal to that expected from individual data
variation, the zones represented by the means are con-
sidered to correlate and, by inference, be continuous.
890
TABLE l-POSSIBLE ZONATION INDICES
Well No.1 Well No.2
Depth Perm. Depth Perm.
JftL
(md) Indices (It)
~
Indices
(a) Two-Zone Indices
8801 91 880. 91
0.66 0.43
8802 89 8805 89
*
0.99 0.90
8803 40 8806 60
0.74 0.86
8804 39 8807 62
*
0.18 0.92
8805 41 8808 29
0.0 0.47
8806 40 8809 28
(b) Three-Zone Indices
8801 91 8804 91
0.99 0.89
8802 89 8805 89
*
0.99
8803 .. 0 8806 60
0.99 0.87
8804 39 8807 62
.
0.99
8805 41 8808 29
0.99 0.80
8806 40 8809 28
{c) Four-Zone Indices
8801 91 8804 91
0.99
8802 89 8805 89
*
8803 40 8806 60
0.99
8804 39 8807 62
.
8805 41 8808 29
0.99
8806 40 8809 28
*Best zonation from two-zone calculation.
**8est zonation from three-zone calculation.
TABLE 2-FINAL ZONATION FOR SAMPLE DATA
Well No.1 Zone Mean Well No 2 Zone Mean
---
91
}
90
91
1
90
89 89
40
}
60
61
39
40
62
41 29
40 Q3
28.5
This mathematical statement is given by Eq. 4:
(kh - k,J> - -+- sz(v.p)'
- - ~ 1 ( 1 1)
2 n
h
n,
(4)
where r:. = the arithmetic average of the permeability
data of the hth zone in one well and k.. = the arithmetic
average of the permeability data of the ith zone in an ad-
joining well, n
h
and n, = the number of data in the hth
and ith zones, s = the standard deviationS of all the per-
meability data from the reservoir and z = a constant tab-
ulated as a function of the number of data, the number
of zones and a probability level. v and p are used to
identify z-values as functions of the probability level.
Harter' provides a table of z-values.
If the left side of Eq. 4 is larger than the right side,
the zones represented by the two means are considered,
on the basis of statistics, to be different. However, if the
left side of Eq. 4 is smaller than the right side, the zones
correlate and can be considered to be continuous. For
example, in Table 2 the top zone of Well 1 has an arith-
metic average permeability of 90 md. The top zone of
Well 2 also has a permeability average of 90 md. If the
two averages are compared by Eq. 4, we naturally find
no significant difference. On the other hand, if the 40-md
average for the second zone of Well 1 and the 61-md
average for the second zone in Well 2 are compared, we
find a difference of 21 md. While we know that these
well-zone averages are not alike, we would like to find
JOURNAL OF PETROLEUM TECHNOLOGY
out if the difference of 21 md is due to different strata
_ or to normal sampling fluctuation.
The difference of 21 md cannot be attributed to sam-
pling fluctuation because it is greater than the right side
of Eq. 4,* which has a value of 2.2 md. Therefore, the
samples probably came from different reservoir zones.
When the mean of the second zone in Well 1 is compared
with the mean of the third zone in Well 2, the difference
of 11.5 md is greater than the right side of Eq. 4, which
has a value of 2.2 md. Thus, neither Zone 2 nor Zone 3
in Well 2 is likely to be continuous with Zone 2 in Well l.
In Table 2, a single 90-md reservoir zone at the top
of the sand is apparent. Statistically, this is all we can say.
Practically, however, we could group the bottom two
zones in Well 2 into one zone with an average perme-
ability of 44.7 md and use this to correlate with Zone 2
of Well 1. We emphasize this "practical" point in order
to show how engineering judgment is used with results
from the statistical zonation technique.
ACKNOWLEDGMENT
The author wishes to express his appreciation to the
management of Jersey Production Research Co. for per-
mission to publish this paper. Acknowledgment is also
due B. T. Willman for his assistance in the preparation
of this paper, and to F. A. Graybill for his ideas and
work in the early development of statistical zonation.
REFERENCES
1. Beghtol, LeRoy A.: "A Statistical Approach to the Zonation
of a Petroleum Reservoir", Master's Thesis, Missouri School
of Mines (1958).
2. Duncan, D. B.: "Multiple Range and Multiple F Tests",
Biometrics (1955) II, 1.
3. Fisher, W. D.: "On Grouping for Maximum Homogeneity",
lour. of Am. Stat. Assn. (1958) 53, No. 284.
4. Graybill, F. A.: An Introduction to Linear Statistical Hypo
thesis (1961) 257.
5. Harter, H. L.: "Critical Values for Duncan's New Multiple
Range Test", Biometrics (1960) 16,671.
6. Kramer, C. Y.: "Extension of Multiple Range Tests to Group
Means with Unequal Numbers of Replications", Biometrics
(1956) 12, 307.
APPENDIX
AN EXAMPLE OF STATISTICAL ZONATION
Fig. A-I shows the location of four wells selected to
illustrate the use of the statistical zonation technique.
The data chosen for the illustration are permeabilities
from a consolidated sandstone reservoir. Table A-I lists
the permeability data and their corresponding depth for
each of the four wells.
STEP I-ZONATION OF INDIVIDUAL WELLS
To illustrate the calculations, we selected the data in
Well 11 because it has less data than the other wells.
Table A-2 illustrates the division of the data into two
zones (the asterisk marks the point of division in Tables
A-2, A-3 and A-4).
We use Eqs. 1 and 2 in the following more-convenient
form for the computations.
1 [L __ ]
B = L _ 1 i: 1 m,(k .. - k. Y
See the Appendix for the calculation of EQ. 4.
AUGUST, 1962
WELL NO. 11

N
WELL NO.8 WELL NO. 18

I_ [000 fT _I
WELL NO. 37

FIG. AI-LoCATw:-; OF WELLS.
1 [L ( . ~ ' k,;)' ( . ~ . ~ i k
i j
)' 1
= __ ~ 3-1 _ t-13-1
L - 1 i = 1 mi N
(5)
w = N ~ L [. ~ .1' (k'j - ki)']
t=13=1
1 [L 11" L ( . n ~ kij r 1
= ___ ~ ~ k';j - ~ --=-3 _-_1_-,--_
N - L i c ~ 1 j = 1 i = 1 mi
(6)
Eqs. 1, 2 and 3 are computed for each division into
two zones and tabulated in the columns labeled B, Wand
R. Note that any negative values in R are replaced by
zero in order to conform to the definition of R. For ex-
ample, the first line in Table A-2 is computed as follows.
TABLE AI-RESERVOIR PERMEABILITY DATA
Well No.8 'Well No. 11 Well No. 18 Well No. 37
Depth Perm. Depth Perm. Depth ferm. Depth 'Perm.
.....J!!L
(md)
-...i!!L..
(md)
~
(md)
~
(md)
1917.5' 11 1906.5' 10 1973.5* 20 1922.5' 34
1918.5 27 1907.5 52 1974.5 40 1923.5 67
1919.5 157 1908.5 276 1975.5 190 1924.5 20
1920.5 234 1909.5 140 1976.5 146 1925.5 197
1921.5 390 1910.5 139 1977.5 53 1926.5 186
1922.5 110 1911.5 156 1978.5 4.8 1927.5 33
1923.5 192 1912.5 342 1979.5 0.0 1928.5 30
1924.5 218 1913.5 87 1980.5 45 1929.5 21
1925.5 42 1914.5 0.0 1981.5 14 1930.5 117
1926.5 120 1915.5 0.0 1982.5 0.0 1931.5 27
1927.5 158 1983.5 84 1932.5 27
1928.5 316 1984.5 28 1933.5 26
1929.5 20 1985.5 0.0 A934.5 61
1930.5 99 1986.5 0.0
1931.5 121 1987.5 0.0
1932.5 43 1988.5 0.0
1933.5 68
1934.5 7.4
1935.5 149
1936.5 0.0
*Top of productive interval.
TABLE A2-DIVISION OF DATA INTO TWO ZONES
Cum. Grand Sum
Sample Sum of Minus
No. Per lPerm. IPerm. Cum. Su.m B W
~
(md) (md)
~ ~
(md
2
) R
'0 10 1,192 13,493 13,600 0
2 52 62 1,140 19,892 12,800 0.35
3 276 338 864 243 15,256 0
4 140 478 724 3 15,286 0
5 139 617 585 102 15,273 0
6 156 773 429 1,118 15,146 0
7 342 1,115 87 35,646 10,830 0.69
8 87 1,202 0 36,120 10,771 0.70
*
9 0 1,202 0 16,053 13,280 0.17
10 0 1,202 0
1,202 (Grand Sum)
891
Eq. 1:
B = [(10)' + (1192)' _ (1202)']
1 1 9 10 13,493.
Eq.2:
W = M (lOr + (52)' + (276), + (l40), + (139)'
(
10)' (1192)']
+ (342)' + (87)' + (0)' + (0)' -
1 9
13,600.
Eq.3:
R = 13,493 - 13,600 = -0008
13,493 .,
which is replaced by 0; and the other lines are computed
in the same way.
The best division into two zones occurs after the perme-
ability value 87 as shown in Table A-2.
Since we have already separated the original data into
two groups, our problem is now that of separating either
Group 1 or Group 2 into two additional groups. Table
A-3 gives details of the calculations.
The second point of division (the first was after the
value 87) occurs after the value 52, and defines our data
as a set of three groups, or zones.
The results of the calculations for the division of the
data into four zones (not given) indicate that the largest
four-zone index, 0.79, is smaller than the three-zone index
of 0.81. Therefore, the well is better described as three
zones.
The other wells were divided into zones in the same
manner, although the details are not given in this paper.
The final results of our work are shown in Table A-4,
where the asterisks mark the divisions into zones. The
identification for the zones in Table A-4 gives the zone
number within the specified well. Thus, Zone (2, 11) is
Zone 2 in Well 11.
STEP II-CORRELATION OF WELL ZONES
On the basis of primary pressure and production-rate
histories for the four wells, there is every reason to believe
the formation is continuous over the area under study.
The second step of the statistical zonation technique is
the correlation of zones in one well with zones in adjacent
wells to determine the size and location of the continuous
zones where they intercept the areas. Each of the four
wells has been divided into zones.
TABLE A3-DIVISION OF DATA INTO THREE ZONES
Sample Grand Sum
No. Per Perm. Cum. Sum IMinus Cum. B W
Gro'lJp {md) (md) Sum (md) (md') (md') R
Group 1
1 10 10 1,192 29,300 9,098 0.68
2 52 62 1,140 37,021 6,893 0.81
.
3 276 338 864 21,450 11 ,341 0.47
4 140 478 724 21,842 11 ,229 0.48
5 139 617 585 22,866 10,937 0.52
6 156 773 429 23,564 10,737 0.54
7 342 1,115 87 20,346 11,657 0.42
8 87 1,202 0
Group 2
0.0 0 0 18,060 12,310 0.31
2 0.0 0 0
H92
TABLE A-4-FINAL ZONATION OF RESERVOIR PERMEABILITY DATA
Well No.8 We'! No. 11 We'l No. 18 Well No. 37
Perm. Perm. Perm. Perm.
Zone (md) lone (md) ZOl"e (md) Zone (md)
11 10 20 34
(1,8) (1,11) (1,18) {1, 37)
27 52 40 67
20
157 276 190
234 140 (2, 18) 197
390 139 146 (2,37)
90 (2, 11) 186
192 156
(2,8) 342
218 87 53
42 4.8
33
120
0.0
158
45
21
316 0.0 14
117
13, i 1)
(3, 18) 0.0 (3, 37)
0.0 20
84 27
20
28 26
99
0.0
61
121
0.0
43
0.0
(3. 8)
0.0
88
7.4
149
0.0
Eq. 4 is applied to the zone data in Table A-4. The
following is suggested as a convenient and efficient man-
ner of applying Eq. 4.
1. Rank well-zone means in order of decreasing mag-
nitude, as follows.
Zone
(2,8)
(2,37)
(2,11)
(2,18)
(3,8)
(3,37)
(1,37)
(1,11)
(1,18
(3.18)
(1,8)
(3,11)
WELL NO. I [
Number of
Zone Mean (md) Data in Zone
192 10
191 2
190 6
168 2
66 8
42 8
40 3
31 2
30 2
19 12
19 2
0 2
WELL NO.8 WELL NO. 37 WELL NO. 18
"}------c")-----,-
ZONE I
ZONE II
ZONE III
ZONE I
.
", ..
.',-
c :.'
. ZONEm
>,
. :.
: ...
.... ,
/
: . ':
AVERAGE THICKNESS (FT) AVERAGE PERMEABILITY (MO)'
2.1
5,0
7.5
33
189
36
FIG. A2-CROSS-SECTION SHOWING FINAL ZONATION.
JOURNAL OF PETROLEUM TECHNOLOGY
2. Calculate Eq. 2 using all permeability data in the
entire reservoir.
W = [(11)" + (27)" + (157)" + ... + (61)"
_ _ ... _ = 3,964 (md").
3. Calculate the standard deviation from Step 2.
s = y3,964 = 62.96 (md).
4. Select z-values' for a 95 per cent probability level
(zv,p)'
p 2 3 4 5 6 7 S 9 10 II 12
Z47,p 2.S6 3.01 3.10 3.17 3.22 3.27 3.30 3.33 3.35 3.37 3.39
5. Multiply the z-values in Step 4 by the standard devia-
tion in Step 3, e.g., F'p = SZ47,P'
P 2 3 4 5 6 7 S 9 10 II 12
F'p ISO 190 195 200 203 206 20S 210 211 212 213
6. Test for significant differences among well-zone
means. First the largest mean is compared with each of
the smaller means. In order for the means of Zones (2,8)
and (3,11) to be significantly different,
(192-0) J 2(10) (2) = 350
" 10+2
must exceed F'12 = 213. It does; therefore, the well Zones
(2,8) and (3,11) represented by the means 192 and 0
are significantly different.
In order for the means of Zones (2,8) and (1,8) to be
significantly different,
(192-19) J2(10) (2) = 316
" 10+2
AUGUST, 1962
must exceed F'n = 212. Since it does, the well zones (or
the means) are significantly different.
Completing the test, we compare the mean of Zone
(2,8) with the mean of the other zones until the differ-
ence is not significant. We find the comparisons are sig-
nificant until Zone (2,18) is reached. At this point, we
begin to compare the next largest mean with the smaller
mean as follows.
(191-0) = 191 (1.414) = 270>F'1l
and thus is significant, and so forth, until a nonsignificant
difference,
(191-168) = 23(1.414) = 32<F'3,
is found.
Finally, we can distinguish the means which are not
significantly different as follows.
(2,S) (2,37) (2, II) (2, IS) (3,S) (3,37) (1,37) (I, II) (I, IS) (3, IS) (3,11)
192 191 190 16S 66 42 40 31 30 19 (I,S) 0
Any two means not underscored by the same line are
significantly different, and any two means underscored by
the same line are not significantly different, as defined
before.
These results and the information given in Table A-4
show that three reservoir zones have been defined as a
middle zone of high permeability bordered on either side
by zones of lower permeability. The properties of these
reservoir zones are summarized in Fig. A-2. ***
893

You might also like