You are on page 1of 10

SPE

mof~

SPE7921

13@_@mac#AlME

GASOCCURRENCE
1N THEDEVONIAN
SHALE

by E.C. Smith, S.P. Cremean, and G. Kozair,


Columbia Gas System Service Corp.

Copyri ht 1979, American Institute


of Mining, Metallurgical
and Petroleum Engirv?ers,
Inc.
This paper was presented
rmeabi f it y Gas Reservoirs,
B&y 20-22,
1979. Denvar, Colorado.
The material
le subject
to correction
by the author.
an abstract
of not more than 300 words,
Write;
6200 N. Centra~+Expy.,
Oallas,
Texaa 75206.

at the 1979 SPE Symposium on LowPermission


to copy is restricted

ABSTRACT
Estimates of recoverable reserves in the
eastern Devonian shales range from a few trillion
cubic feet to several hundred trillion cubic feet.
The most pessimistic estimates assume that all
recoverable gas occurs within natural fracture
porosity. More optimistic estimates assume a
substantial gas contribution from the shale matrix
close to fractures. This paper presents the available evidence for both views including long-term
shale production characteristics,laboratory
measurements on shale cores, and numerical modeling
studies. The weight of all available evidence
favors the position that the matrix provides a
major contribution of the recoverable gas from the
Devonian shales, with a smaller amount derived from
the fracture void volume.

what fraction can be produced economically within


a reasonable time frame. Since the shale has a
very low permeability,economic production requires
a fracture network to provide sufficient flow,
except whece the secticn sontains permeable siltstones or sandstones. Gas production rates wi11
depend upon the s!ze, spacing and geometry of these
fractures.
The extent of these controls is the subject of
a lively debate which focuses on the dynamics of
shale gas production. Two basic schools of thought
a) the balk of the recoverable gas derives
exiW:
from the matrix, but fractures are necessary to
provide rapid flow to the wel1, and b) the bulk of
the recoverable gas is contained within natural
fractures in the shale section and the contribution
from the matrix over the life of the well is unimportant.

INTRODUCTION
For several years, the Columbia Gas System
Service Corporation has shown an increasing interest
in promoting Devonian shale research and development
t
in the Appalachian Basin. A systematic exploitation
program would hinge upon the drilling ofa large
number of test wells in key parts of the Basin.
This, in turn, requires prior demonstrationof the
basins potential gas supply. Our research cantinues to support estimates of 200 to 900 trillion
cubic feet of potential gas supplies based on production records, core gas content measurements,
well logs and well cuttings descriptions.

Several reports published by the Office of


Technology Assessment, Federal Power Cormnission,
Federal Energy Regulatory Cormnission,TRW and Lewin
and Associates on the Appalachian Devonian shale
provide other resource estimates using different
assumptions. One very comprehensive report by the
National Petroleum Council should he issued in the
-near future.
/
Althoughwe know that the shale matrix contains
a large volume of mobile gas, the question remains
99==

This debate on fracture vs. matrix production


began at least 45years ago. The proceedings @f a
1935 Devonian shale symposium mentioned both concepts.2g) Each concept claims the support of production data and history matchifigof production
decline curves. Both sides can account for the
total gas production (using appropriate assumptions
and input data). Both have the support ofnumerical models.
Columbias interest in these production models
stems from their influence on reserve estimations.
We must resolve whether matrix gas can be recovered
before using the matrix gas content as a basis for
supply projections. This paper reviews the assumptions, input data, and calculations used in each
approach. Although we may not definitely prove
which is correct until after a widespread drilling
program, we can compare the likelihood of each concept. Comparing the two arguments may help us
narrow the range of values for the assumed input
data and gain some insight into fracture size and
spacing. These factors should prove useful in
determining recovery factors, well spacing and
fracture designs.

?,

HISTORICAL INFORMATION ON SHALE GAS PRODUCTION

Devonian shale exploitation has endured since


1820 when the first gas well was drilled in
Fredonia, New York. By 1890, numerous very shallow
wells were drilled from Dunkirk, New York to
Sandusky, Ohio. During this century, mo!jtshale
production came from eastern Kentucky, western
West Virginia and southern Ohio. To date, development has concentrated on only 4,000 of the over
110,000 square miles of the Appalachian Basin under
lain by tle shale.
In .he earlier years, very shallow wells with
pressurss barely above atmospheric provided an
easy .S{JUrc(?
of small volumes of gas for local use.
From 1927-1962, th[!economic climate encouraged
sha?e gas exploration as can be seen from the
number of completed wells by Columbia (See
Figure 1). However, the industry typically
explored for other targets, and in some cases stimulated the shale if the other targets failed to
produce. We know of large shale gas shows which
interrupted drilling, but which were not exploited.
Production records(lG) show that 40% of the
Shale wells had no measurable natural show and
only 5% had shows of a commercial size (Figure 2).
Most gas shows occur within the black, organicrich section which contains an avera e of six tim s
as much matrix gas as the remaining 7gray) shales?6
After treatment with gelled nitroglycerine,89% of
the wells have produced gas commercially. An additionc~ fraction of the firywells Woi:ld have produced gas but for mechanical problems. Past well
spacing practices reflect no syst(:maticmethod, but
show the effects of 1) competitive drilling with a
high well concentration near lease boundaries,
2) less drilling near dry holes, 3) availabilityof
proper land titles, 4) topography and cult
features, and 5) company spacing policies.yf!
Consequently,on? must question the accuracy of
drainage volumes, recovery factors or other
calculationswhich assume that well spacing equals
the drainage radius.
During the past six years, Columbia and
several tens of research, service and supporting
companies have studied the characteristicsof the
Devonian shale, recently in cooperation with the
U.S. Department of Energy. Several of these institutions have provided data which will help us compare the validity of the two gas prodl!:i.ion
concepts.

=%ib)w~-it)

(1)

2) wherein fractures are so closely-spacedthal


the drainage zones of adjacent fractures effectivel~
overlap or interfere. This relation describes the
production of very high volume wells:

q=-+-b

(2)

Kuuskraa et al 8) claimed that diffusion from


the shale proceeds too slowly to account for the
observed production. They claimed that Schettlers
sample preparationmethods (i.e., crushing and
drying) could have altered his measured sorption/
desorption rates yielding an erroneously high diffu.
sion constant. The constant depends upon the radius
or thickness of sorbing grains or slabs. They
rcesoned that micro-cracks and laminae se arations
(assumed to exist in the prepared samples! could
increase the value of the constant.
However, Schettlers earlier studies(11) show
that the particle size has a small effect upon the
diffusion constant. In fact. Schettlers model
already considered this effect. Columbias off gas
rates measured on unprepared,whole core samples
lend further support to the dffusion release
rates, or show the constant to be conservative.
Kuuskraa et al(8) voiced a second objection to
Schettlers model. They contend that a log-log
plot of fractional gas release versus time should
have a~slopeof O.5 for a diffusion-controlled
process until halfof the gas has been produced.
They reject diffusion as a controlling process
based on observed initial slopes of 0.8 to l.O which
decline to 0.4 to 0.6 over 20 years.
However, this assumed production slope ofO.5
corresponds to d+ffus?on from the surface of a semi.
infinite slab releasing gac in response to a sudden
pressure drop to a constant pressure (3 and 7). In
contrast, Schettlers flow constrictions imply that
the pressure in the fracture syctem does not
suddenly drop to a constant pressure. Carslaw and
Jaeger(3) provide equations which describe diffusior
flow under more realistic pressure-flowconditions .
i.e., for gas release related to a linear pressure
drop versus time with a similar production slope of
1.5, or for a constant flux rate at a fracture
surf~-ewith a slope of 1.0(7~.

DISCUSSION
Matrix Gas Model: Schettler(l1) derived a numerics
model from Ficks first and second laws of
diffusion(5)which generated a successful history
mat:h of actual shale gas production for high,
medium, and low volume wells.(Figures 3 and 4).
The model considers two separate cases:
1) wherein constrictions in.the fracture network (i.e., mineral deposits or partial closure)
limit the flow rate of gas diffusing from the
matrix into widely-spaced fractures. The correspending equation accounts for all but the highest
production rates:

Elkins(7) estimates an expected production


slope of slightl.vless than 1.0, provided either
diffusion or-Dar~y flow accounts for matrix gas
release into the fracture system. He reasoned that
the gas flow rate in the fracture system controls
both the pressure drop in the reservoir and the dror
between the fracture surfaces and the wellhead. The
pressure drops in turn, determine the production
rate. Matrix gas release increases as the pressure
in the fracture system drops, but the fracture flow
capacity decreases. This analysis made use of the
following three relationships:

The resulting porosity and permeabilityvalues


range from 0.31 to 0,8% and fr~m 0.0175 to 0.027
red.,respectively.

For a stepwise linear pressure decrease in the


fracture system:
Pt=

Pi-mot-ml(t-tl)-...mntn)tn)

(3)

..
For estimating stepwise gas.influx to the fracture system whicn decreases linearly with time:
%

= rmot15+rm1(t-t1)15+....rmntn)1515

(4)

For the average gas production during each


step - based on the back pressure gas flow
equation:
.
.K~:-p:+:t-l-p:)

(t(t-,))

(5,

t-%-l
By considering the fracture porosity small compared to the gas storage capacity of the matrix,
he didnt need to Involve pressure transients in
the fractures. Elkins initially assumedr= 1.0,
P. = 50Gpsi, and P = 490 psi. The above solut]on holds until th~ production from adjacent
fractures begins to interfere - i.e., until a
large fraction of the gas content has drained from
the matrix.
Schettler has shown that the rate of gas diffusion from the Devonian shale matrix can satisfactorily explain the observed production rates.
The objections to the expected slope of the
productivity-timedecline and the diffusion
coefficient do not appear warranted.
Fracture Porosity Model: According to the fracture porosit.ymodel, virtually .Illproducible
gas resides within the naturally-occurringfracture system. Release from the shale matrix proceeds too slowly to contribute significantly.
The support for this model rests on a history
match of production data with calculated Dai-;y
flow prod~~~ion in a uniformly fractured
reservoir~8]. A fracture porosity and permeability so derived can account for the produced
gas volumes. However, a reviewof the input data,
assumptions and data handling provides further
insight into the problem. The matched production
data consists of averages for several groups:
high, medium and low production wells plus
averages over separate geographical areas. After
smoothing the averaged data with Marquardts
Algorithm to fit the curve defined by
f(t) = (l-e-Bt)

(6)

Kuuskraa et al(8) separated the data from the


resulting curves into annual production after
5, 10, 20 and 30 years. Figure 5 shows the
resulting history match for cumulative prod!!ction
after 5, 10, 20 and 30years based on the fjllowing assumptions:
Well spacing = 150 acres
Gas gravity= 0.6
Net formation thickness = 580 feet
Reservoir temperature = 560 R
Initial reservoir pressure= 500 psia
Flowing bottom hole pressure = 100 psi

Certain aspects of the history mat~h merit


further attention. Figure 5 shot the history
match for several cases. First, note that only fou
data points define each production curve. Each
point represents an average over a time i~terval
for a number of wells.
After fitting the original production data to
an equation with Marquardts Algorithm, the shape
of the resulting curve differs from the original
datas trend. Figure 6 compares the shape ofactua
production decline data (averaged for groups of
shale wells) before and after this smoothing
process. Considering the distinct differences in
the curve shapes, valid history matches should
probably use unsoothed data and compare more than
four points along a curve.
Eikins(7) and others
have questioned the use of this smoothing function.
As shown by Figure6, equation 6 imposes a constant
percentage decline Pate on the data. This yields a
linear production rate vs. time on semi-log paper.
A valid smoothing procedure would use an algorithm
which accords with the observed production behavior
Proponents of the fracture production model
claim that the fracture network can contain the
volume of shale gas produced during a wells first
30 years(8). This assumes that al1 of the gas
occurs within a 500 foot netpay thickness, over a
150 acre area with a fracture porosity of 0.3 to
0.8%. Actually, certain areas would require in
excess of 4% fracture porosity. For instance, one
part of Floyd Co:mty, Kentucky drilled on less than
a 40 acre spacing averaged better than 100MMcf per
wel1. Using the recovery factorof 65% used by
Kuuskraa et al8) and the following equation:
=

AhTs@Pf

(7)

zfps
defines a required porosity of about 4%. (Remember
that the many variables which have determined the
well spacing make it risky to estimate the true
shale volume d~afnedby a given well). In either
case, these required fracture porosities seem much
higher than those indicated by shale core samples.
Core analyses and descriptions ofover 2500
feet of Devonian shale from producing areas of the
Appalachian Basin (Lincoln County, West Virginia
and Martin County, Kentucky) suggest porosities of
0.0003% to a maximum of 0.01%. (This is based on
observations of less than one natural fracture per
feet and fracture widths frcm 0.001 to 0.03 cm as
measured by Terra Tek of Salt Lake City6)). Moreover, the effective fracture porosity may not reach
these levels as a result of observed carbonate
fracture filllngs.
Columbia has reported effective permeabilities and frac lengths calc:llatedfrom transient(6)
reservoir tests in two West Virginia shale wells
The test interpretationscome from several sources
using a variety of assumed flow periods (ideal
linear, radial and spherical flow). Because the
testing took place during the post-frac clean-up

14.

GAS
OCCURRENCE
IN THE
------....-. .- -..
. .. DEVONIAN SHALE
process, we used the results to compare the effects
of our fracture treatments, but did not-use the

results quantitatively. Kuuskraa etal(8) mentioned that our reported permeabilities lend
support to those generated by their history match.
t
Elkins(7) questioned the basis of this support
We found that our proppant volumes and calculated
fracture lengths require fracture widths of 0.7 to
11.5 inches. In contrast, Halliburtons rock
mechanic$ specialists suggest a maximum width of
0.5 inches and a probi~bleaverage width closer to
0.1 inch. Due to the tests short product~on(71
periods compared to the build-up time, ElkIns
rejects the use of type curve matching for tietermining permeabilities,frac+tirelengths or for
demonstrating radial versus linear flow. Because
of the test period lengths, resulting log-log plots
of build-up pressure vs. time may display a false
linearity(Figure 7).
Elkins(7) noted a linear relationship between
P2 and (-@
for extended times in all of
our tests. Taking this as an indication of linear
flow from the matrix into the induced fractures
(O.1 fracture width with 25% porosity), he calculated an zffective permeabil::yrange from 0.8 to
3.2 microdarcies. This calculation employed the
following equation for linear flow of constant
compressibilityfluids from a semi-infinite slab.

The matrix porosity (0.7%) was determined from a


projected ultimate recovery of 575 MMcf per well,
a 500-foot target section, a 150-acrewell spacing
and a 400 psi average pressure drop.
In view of the data smoothing and curve fitting procedures, the present fracture porosity
history match does notappear valid. Moreover,
this model requires porosities and permeabilities
at least one order of magnitude higher than those
indicated by core analyses and the most careful
interpretationof our reservoir tests. Consequently, this model does not adequately support
production from fracture porosity alone.

SPE 7921

matrix gas content ofa volume of shale 500 feet


thick over a 150-acre drainage area. Their matrix
gas content ofO.21 cubic feet of gas per cubic
foot of shale reflects Columbias reported average
for the entire Devonian shale section in an area
which appears to be half depleted. In ar?ycase,
we found that the shale within the 500 foot stimulated zone now contains about three times the
average content. This would suggest a fracture
frequencyof less tharlone fracture per200 cm.
ElkinsS(7) estimate of the spacing required
for production from the matrix derives froma
Car;law and Jaegers(3) equation:
n.m

~p.mt+!M&Li+,.m@~
Ed..
n=o

KW3
(9)

@.)!!-e
(2n+l)3

-K(2n+l)2r2t/4L2
. @s ~
:,

.X

This re?~tionship holds for linear fluid flow from i


slab into a fracture where pressure decreases
lir,earlywith time. These results suggest a fract~re spacing of 1000 feet for a 0.57 microdarcy
permeability (where m= 0.0488psi/day for19441960, and m = 0.0422 psi/day for 1960-1975).
Elkins plotted the cumulative production vs. P/z fo
38 Columbia Gas shale wells in Lincoln County, West
Virginia. He noted linear pressure drops from 504
to219 psi during 1944-1960 and from219 to 183
during 1960-1975- the two slopes reflecting
different historical development trends.
Compared to the matrix model, the fracture
poroAity model demands a greater fracture frequency
Using Muskats equations and the porosity and permeability defi~~$ by the fracture volume model,
Kuuskraa et al~u) found a required spacing of
5 X 10-3 cmwith fracture widths of2.5
,011W3
k

Fand

X 10-5 cm.

w
T=+

(10,11)

Such a close spacing contrasts s~f~ply with th


A Comparison of the Two Models: As a further test
of each models practicality,we can review the
fracture volume and spacing requirements: the permeability and porosity required for the fracture
porosity model; and the expected gas flux from the
shale matrix. AL this point, field data, core
descriptions,analyses and material balance curves
provide useful insights into the problem.
8oth production models agree that the fracture
spacing controls or limits shale gas recovery rates. Schettler12) f,~nd that a single fracture
of indefinite lateral extent with a width of 0.01
cm which intersects 40 meters of wellbore could
prov?de a flow of 50 Mcfd after one year. Some
readars may object to the use of frac;~~es of
indefinite extent. However, Kuuskraa~u] calculated
that the diffusion model requires a fracture
spacing (2L) of 100 cm. They determined this
using a plot ofnnrmalized time
Paci!y
(kt/L versus the recovery fraction(Figure 8).
Their estimate of the recovery efficiency (O.43)
derives from the 30-year productiondividedby the

actual frequency of natural fractures~o~ observed i


cores from the producing area (up to one fracture
per10 feet). If the spacing were sufficient to
permit the fracture porosity model, the shale cores
should have much higher penneabilities because they
must each contain numerous fractures (or conversely
theeffect~ve permeabilitymust be much too high).
Fracture spacings which do allow for significant
fracture porosity production have a much greater
potential for matrix production due to the enormous
exposed surface area.
According to the fracture poros~ty model, flow
from the shale matrix amounts to virtually nothing
over a 30year well life. In order to quantify the
matrix and fracture contributionsover any time
interval,we would need to knew the actual fracture
spacings and the pressure profile in the fracture
network. However, we can find the time required
to refill a fracture volume per unit area of
fracture for any given pressure drop AP as follows:

For flow by diffusion from both fracture surfaces: I

time for the present reservojr:

=(*JandG=(!fwi~i)
12$13
t=*

(15)

(The last term in equation 13 derives from equation

Usfng the followfng fnput data, he ffnds a requfred

7). Along 1 cm of fracture, the maximum width


should be .03 cm accordfntato Terra Teks core

tfme of 3.4 years: A= 150 acres, c= 5X 10-3psf-\


u = 0.012 centipoise, $ = 04,005,k = 0.025 mfllfdarcy. Ifwe belfeve thfs equatfon applfes and the
assumed data fs correct, we would have to concede or
thfs point.

analyses(6). Based on vaiues of d = 2 X 10-7 for


Lincoln County, West Vfrginia core12), flow from
the matrfx could reffll the fracture fn 0.65 hours
foran.y pressure drop no matter what fractfon of
the fracture gas fs produced.
Thus far, we have only considered diffusion as
the process of matrix gas release. Hovever, for
lfnear Darcy flow fnto a one square-foot fracture,
the time required to recharge the fracture fs:
t=~96X103GTz
~
~
kc+
A(Pf -Pt )

(14)

(derfved by integratingequation 8 with respect to


time). Then, for pressure drops in the fractures
rangfng from 10 to 400 psf, the recharge tfme
ranges from 6.8 X 10-4 to 1.9 x 10-3 hours (roughly
2-7 seconds).
Actually, the pressure doesnt drop instantaneously throughout the fracture system. However, we
may consfder the matrfx to respond instantaneously
to the actual pressure drop anywhere fn the system.
Therefore, whether by Darcy flowor diffusive flux,
the matrix flow capacity as a whole exceeds the
fracture f1ow capacfty as a whole. From the
ta
on the matr!x gas content and fracture width($ , we
also know that the matrix wfthfn one foot of a
fracture contains at least 40 times that fn the
fracture volume.
While the above recharge tfmes are calculated
estimates rather than observed rates, Columbia has
measured offgassing rates for thfrty core samples.
These rates are slower than those expected by
Linear Darcy flow but faster than those calculated
by the diffusfve flux (for gas release to atmospheric pressure). Our average off gas rate was
0.0013 cfgas per square footof surface area per
day at 70F. However, the fnitfal rates exceeded
ten tfmes thfs value. Also preliminary data at
elevated temperatures suggest that the rati2may
fncrease more than an order ofmagnftude at formation temperaturesover 90F.
Another lfneof evidence which suggests (but
does not prove) a dual porosity is the slope of
n%lterialbalance curves (P/z vs. G). Ffgure 9
shows the slope change during 28 years of production, characteristicof Devonian shale wells. The
classical explanation of thfs observed behavior
would relate the early declfne as dominated by
fracture volume depletfon followed by the more
gradual matrfx depletfon.
Kuuskraa et al 8) have pointed out that the
48 hour rock pressures used to generate these
curves may not be the true volumetric average
reservofr pressure. They refer to the followfng
equation for dete?mfnfng the requfred stabilization

However, a revfew of well pressure records


brings to lfght an intrstfng contradiction. In a
carefully studfed area ncludfng Lfncoln County,
West Vfrgfnia, wells completed primarily between
1947 and 1956, Elkins(7) found that fnitfal 48 hour
build up pressures had depleted values approaching
the annual 48 hour pressures of older producing
wells (Figure 10). The oldest wells had fnitfal
reservofr pressures in excess of 525 psi. Pressures
fn the ffve wells completed in 1956 ranged from 350
to 390 increasing toward the undeveloped area.
Three new test wells completed from 1976-1978
pressured up to about 250 psi. All this strongly
suggests an interconnectedfracture network and
that areas well above 150 acres can be affected
wfthln 30 years. Elkfns calculated an effectfve
permeability of O.004md from the data for the 1956
wells whtch flowed 30 Mcfd for three weeks before
their first48 hour bufld up (using a flowing BHP of
about 100 psia at an average formatfon pressure of
400 psfa, and assumfng a homogeneous formation wfth
0.7%gas fflled porosity, an effective well radfus
of four feet and 500 feet of net pay).
Ifwe considered an unsteady state radfal flow
(as fn the fracture porosfty model), the drawdown
half way between arrew well and anefghboring old
well produced for seven yea~s should be only 10 psf,
In contrast, the initial 48 hour pressure of the
three new wells (up to 4000 feet from the oldwells~
were wfthfn 5 to 18 psf of the annual measurements
fn the nearby older wells. It appears that the
pressure trrthe fracture system can equalfze over
more than a thousand feet withfn days.
This again suggests an extensive fracture
system wfth a very low flow capacfty drafning a
large volume ofmatrixwfth a greater potential for
gas release.
Ifmatrfx gas release accounts for most shale
gas production, the requfred fracture concentration
is four orders of magnitude less than that necessary for flow solely from fracture porosfty.
Actual fracture spacings observed fn shale cores
compare with those required for gas production from
the matrfx. In additfon, the shape of the G vs. P/i
plots, and the simflar bufld-up pressures fnold
and recently completed wells suggest a dual porosit~
and an interconnectedfracture system drainfng the
shale matrfx over a large area. Perhaps the most
convincing evidence formatrfx gas production fs the
rapid rate at which matrfx offgassfng can reffll a
fracture volume after the removal of any portfon of
the gas. This sfmple feature fndfcates that matrix
gas release assumes a dominant role early in the
lffe of a Devonian shale.

CONCLUS1ONS

NOMENCLATURE

A reviewof the proposed models which purport


to@xplafn Devonian shale gas production revealed
thatthe matrix gas release rate into the existing
fracture system exceeds the bulk flow capacity of
the fractures.,,The matrix will begin to contribute
to a wells production as soon as gas begins to
flow in the fractures. Because of the shales gas
content release rate (by either diffusion or Darcy
flow), the matrix must be considered the major
source of produced gas. Consequently,the distribution of matrix gas concentration forms a valid
basis for estimating the potential gas supply in
the Appalachian Basin. However, until we know
the actual fracture spacings, dimensions, flow
capacities and pressure profiles for a given well,
we can not quantify the matrix and fracture porosity contributions accurately.

a = a constant - 7164mcf/day%
b= a constant- 40.5mcf/dy
c = compressibility- psia-f
d= specific degassibility - in cm3/cm2/torr/sec&
f(t) = a function of time
9 constriction faotor - ft
h = net pay thickness
k = permeability - millidarcies
m= pressure decline slope - psia/day
q = production rate - mcf/day
r = a constant that includes effects of fracture
area, permeability, and diffusivity - mcf/psia/

Ifwe consider that all recoverable gas


initially resides within a fracture network, the
required fracture spacing and permeability become
suspiciously high. Core analyses and descriptions
of 2500 feet of core-attest to the error of this
fracture porosity model. In contrast, the parameter values required by matrix gas release fall
within the range actually observed. In addition,
the calculated release rates compare with the
natural offgassing rates measured on shale cores.
Finally, the fracture porosity history match
with shale gas production contains several faults
;~~;;t;ave apparently given rise to the misleading
. So far, the history match based on diffusion appears valid and could provide additional information on fracture geometry, pressure profiles,
drainage volumes, recovery factors and on the general nature of shale gas recovery.
Two other ~pproaches could shed more light on
the factors controlling gas production: 1) Performing long-term tests in closely-spacedshale
wells to define flow conmnication in the fracture
system, and 2) Mining a volume of shale to study
the natural fracture frequency and other relevant
characteristics. TheU.S. Department of Energy is
currently considering both ideas. Off gas analyses
on cores from the depleted area in Floyd County
with a 37-acre well spacing might also be revealing
The present study shows that any numberof
numerical models can explain shale gas production
using convenient data and assumptions. In weighing
the cases for matrix and fracture production, this
review depended heavily upon core data, because
these provide more clear-cut and reproducible
evidence.
Resource estimation is a tricky task at best,
and especially so for the Devonian shale. Consequently, we would warn the reader not to accept
any estimate of Devonian shales gas resources
without a thorough understandingof their basis.
(For this purpose, al1 of the 1isted references
will be helpful). We believe that the forthcoming
report by the National Petroleum Council should
better define the volume of the shales possible
gas.supply.

.-.

s =
t =
w=
x =
z =

day%.
fracture spacing - cm
time - days
fracture width - cm
distance in x direction - ft
gas deviation factor - dimensionless

A=
B =
C =
D =
G=
K=
L =
P =
T =
v=

area - ft2
a constant
gas concentration- cf gas cf shale
diffusion coefficient - fti/day
cumulative gas production in mcf
a constant= 1000 k/(w$c)
fracture spacing + 2 - feet
pressure - psi~
temperature - R
volume - Cf
,

Greek

o = porosity - dimensionless
u = viscosity - Cp
Subscripts
f =
i =
n =
s =
t=
w =
x=

formation
initial
nth term
standard
value at time t
well bore
open fracture

ACKNOWLEDGEMENT

The authors gratefully acknowledge the efforts


of Lincoln F. Elkins of Sohio, Paul D. Schet.ler,
Jr. of Juniata College and Todd M. Doscher of the
Doschers Group in researching the mechanics of shalf
gas production. Most of the mathematical approaches
reviewed in this paper come directly from reports
by and correspondencesbetween these individuals.
REFERENCES

1.

Brown, Porter J.: Energy from Shale -- ALittl~


Used National Resource, In: Symposiumon
Natural Gas from UnconventionalSources,
National Academy of Science, Section II,
Chapter 6, pp. 86-99, (1976).

2.

Browning, IleyB.: Relation of Structure to


Shale Gas Accumulation, In: Devonian Shales A Symposium by the Appalachian Geological
Society, Charleston, Vol. I, pp. 16-20, (1935).

9.

Lafferty, R.C.: Occurrence of Gas in the Devo.


nian Shale, In: Devonian Shales - A Symposium
by the Appalachian Geological Society, Charleston, Vol. I, pp. 14-15, (1935).

3.

Carslaw, H.S. and Jaeger, U,C.: Conduction of


Heat In SolIds, 2nd Ed., 51~ p., Oxford Press,
(1959).

4.

Chase, Robert- P.C. (Professor,Petroleum


Engineering, Marietta College, Ohio).

5.

Crank, J.: The Mathematics of Diffusion,


Oxford Press, London, (1954).

6.

:=;

7.

Elkins, Lincoln: P.C. and conmmnicationson


fractured reservoirs to Todd Doscher of Lewin
and Associates and Don Ward of U.S. Department
of Energy, Washington.

12.

8.

Kuuskraa, Vello A., Brashear, J.P.. Doscher,


ToddM., and Elkins, Lloyd E.: Enhanced
Recovery of UnconventionalGas Sources,
Section on Devonian Shale Gas of the Appalachia
Basin, Vol. III, pp. 1-79, (1978).

13. Smith, E.C.: A Practical Approach to Evaluating Shale Hydrocarbon Potential, In:
Proceedings Second Eastern Gas Shales Symposiuml
Morgantown Energy Technology Center, U.S.
Department of Energy, Morgantown, Vol. II,
pp. 73-87, (1978).

~.P., McKetta, S.F., Owens, G.L. and


. Massive Hydraulic Fracturing
Exper;me~t~ofthe Devonian Shale, Columbia
Gas System Service Corporation, Columbus,
Vols. I and II, (1979).

10.

Ray, Edward O.: Devonian Shale Development in


Eastern Kentucky, In: Symposium on Natural
Gas from UnconventionalSources, National
Acade~ of Sciences, Section II, Chapter 7,
pp. 100-112, (1976).

11. Schettler, Paul D. Jr.:


Studyof Hyiro~arbonShale Interaction, In: Report #ORO-5-197-l
thru 11 for U.S. Department of Energ , Juniata
College, Huntingdon, pp. 1-13, (1976Y .
Schettler, Paul D.Jr.: ATwo Step Model for
Gas Production from Low Permeability Shales,
In: Report #ORO-5197-11,Appendix C, pp. 1-18,
(1978).

r COMMERCIAL OPEN FLOW


I
I

Ftg. 1- Columbtas Devonian shale gas development activl ty.

Fig. 2- Frequency of natural open flows in Davonain s;ale wel1s


wells.

320 -

260
240
t

200
-t

40

12

16
YEAR6

:-

2C

Ffg. 3- History match of shale gas production with gas recovery


calculated by diffusion (for low and intermediate production).
After Schettler.11 )

Fig. 4- History fiatch of shala gas production ISith gae


recovery calculated by diffusion (for high production).
After Schattler.11 )

400. HIGH CASE

Soo.

SMOOTUSD DATA

a H16H 2622
: il$X&d&E

4 HWI CA6E
m Kuw
CASE
LWC5S4

\
\

MEOIUMC&

ACTUAL DATA

6
5
4

300

200

o
LOWCA%E
II

200

60
60

40

=..,,.,

$
g

- /ij
u

,00
ACTUAL

PROOIJCTIOH
MTA

0 Awwiq

1~

20
,0

D Amllwlg da twwltol
fmwwa Prtaolt.

-r-

30

9i14

6Ulbli?

14161Stiti24

262jkwAJlS

&

io
TIME ( TEARS)

Fig. 6- Production decl tne of Devonian zhale WS11s.After Elkfns(7)

match of simulation and ffeld data


Ftg. 5- Hietory
for thres t ypf cal Devonian nel 1s. Af far Kuuakraa
et al,(8)

-100,000
Owlatioh
fromLinwily
0
0

:B
K

-10,000
PR@UCTION PERI~

= 200

,10

Fig. 7- Theoretical prassure but ld up for I fnear flow.

HOURS

,100

After Elkfna{7)

1,000

[{

,4

.3
!
*.e

100

.1

i
o+
o

~Y,,,,,
!1

.3

.4

.5

Nofmolixd

.6

.7

.8

.s

<0

1
200

100

I
SW

I
400

*
C&o

fimo ( Kt/~}

Gp
Ftg. 8- Relationship of recovery efficiency versus time.
Aitsr Kuuskraa et al (e)dsrived from Carelaw and Jaegers
iiork, (3)

(WWF)

Ftg, 9- P/z versus cumulative production for a


Columbia Davonian ehale well.

......
11

... .

y.11

Sg.ii

OA

1 74

AA
qe ~

8 O*

y.,

~ +

176.

fvf

sp.>

spa

sp

Wy

q#.a

.m.8

......................................

~a

-.

w
_..Awmbllec

- -

akmah

u ~Se.s

emmr~-wt

RDmd

Fig. 10-1956 shut-in preesura distribution


in
Lincoln/f!ayne County ehale nel 1s. After Elkins.(7)

You might also like