Вы находитесь на странице: 1из 6

GroundwaJer. Past Achie1.ements and Fw..re ChaJ1enges. Sii1 at (ecs) i; 2C(X}Eakma.

,.. ~

gq 5i5C!JTs:Ji

Droundwater reservoir characterization based on pumping test curve

diagnosis in fractured fonnation
L Bardenhagen
lJeparrmenta/Wo/er Affairs, WindJwek,Namibia

ABSTRACT: In this paper methods and techniques for the diagnosis of pumping tests in fractured aquifers
are reviewed and extended. These techniques, which include the comparison of drawdown and recovery data,
straight-line analysis, special plotS, derivatives, and pseudo-skin analysis are demonstrated on a continuous
fractured model (double porosity case) and a discontinuous model (single fracture case) using theoretical and
practical examples.

Pumping tests are the most important tool for the

aquifer investigation in the groundwater industry, as
they are the only method that provides information
on the hydraulic behavior of the well, and reservoir
boundaries. These are essential information for an
efficient aquifer and well field management mId frequently, togetlter with drilling results, are the only
information available for. tlte establishment of
groundwater extraction schemes. To achieve reliable
reservoir information in fractured aquifers, a sound
understanding of tlte drawdown behavior is of high
importance. A detailed diagnosis of drawdown data
in combination witlt geological data can help to
comprehend tlte reactions in a llactured aquifer. This
paper will briefly review the two important and
well-known models for fractured aquifers, the continuous model for double porosity or natural fraclUred aquifers and the discontinuous model for a
single fracture or fault embedded in a homogenous
matrix. These models will be used to demonslrate
the diagnostic tools available to identify tlte different
flow phases tltat can occur during a pumping test in
. fractured environment. Field data will be presented
to illustrate and discuss some of these cases.

2.1 Double porosity models
The concept of double porosity introduced by
8arenblatt et al. (1960) considers homogenous dislributed conductive fractures embedded in a homogenous distributed matrix. Two matrix types are

. 81

generally discussed: tlte spherical block matri,x

'(Warren & Root 1964) used to represent aquifers
like quartzite and tlte layered matrix (Kazemi 1969)
adopted for example for sandstones with bedding
planes. Commonly a pseudo-steady state block-tofracture flow is assumed, which is a special case of
transient flow reslricted by skin, as illustrated by
Moench (1984).
Moench (1984) considers that two representative
elementary volumes (REV), one for ~ fractu&s
and'one for the matrix, are necessaryJor tlte matltematical description of tlte flow in such aquifers. The
classic concept of double porosity cannot be applied
if tlte cone Qf depression is not considerably larger
tItan botlt REV because tlte influence of tlte single
ftactures is not negligible as demonstrated by Wei et
al. (1998).
The typical drawdown curve in a double porosity
aquifer is shown in Figure 1. Initially, mainly tlte
fracture network releases water and tlte drawdown is
characterized by a straight line in a semi logarithmic
(scmilog) plot. The flattening of the curve is originated by the ever-increasing additional contribution
from the matrix. Thc later drawdown is the response
of tlte matrix alone, which. is represented by a
slraight line parallel to tlte initial one in the semilog
plot. In a double logarithmic (log-log) plot, botlt the
initial and tlte late drawdowns are characterized by
Theis curves shifted horizontally from each oilier.
The distance between tlte two parallels or two Theis
curves depends on .the fracture/matrix storage coefficient ratio.
Both tlte pseudo-steady state and transient models
show an almost horizontal flattening when block-tofissure skin is present. The flattening in tlte transient
model witltout block to fissure skin is not horizontal.


It has a slope of half the slope of the two parallels

(Fig. 1).

Additionally, Cr is used to classify the different

flow phases that can occur while pumping from. a
vertical fracture as follows:
Cr < 0.1, bilinear flow (after well bore storage or
linear fracture flow), thereafter pseudo-radial
- . 0.1 :s Cr < 100, bilinear flow, linear formation


Cr ~ lOO,linear formation flow

When these single vertical fractures are represented in a log-log plot, the following characteristics
can be depicted (Cinco-Ley & Samaniego (198Ia):
Linear fracture flow appears as a straight-line
with slope 0.5 in the very early time data
Bilinear flow is plotted as a straight-line with ...
slope 0.25 in the early time data (Fig. 2)
Linear formation flow shows a str!rightline with
slope 0.5 in the intermediate time data (Fig. 2).
The radial-acting flow phase plots on a straight
line in a semilog plot.


flow appears masked in the transition zone


I (mini

Figure 1. Drawdown
in slab-shaped
double porosity
and transient block to fissure flow, A = fracture/formation
skin-factor ofO.
factor of 5. B = fracture/formation


2.2 Singlefracture models

In Iow permeability rock, features like fractures,
faults, dykes, or bedding plains can represent zones
of high permeability that act as conduits. Basically
five models can be applied for these features:
Finite vertical fracture with infinite fracture conductivity (Gringarten et al. 1974)
Finite vertical fracture with finite fracture conductivity (Cinco-Ley & Samaniego 1978)
Infinite vertical fracture with finite fracture conductivity (Boehmer &
dyke/formation m'6del)
Finite horizontal fracture with infinite conductivity (Gringarten & Ramey 1974)
Finite horizontal fracture with finite conductivity
(Valk6 & Economides 1997)
Whether a vertical feature is finite or infinite
conductive depends on the relative conductivity Cr
defined as:



I (minI
Figure 2, Drawdown
ferent relative conductivity

for vertical


with dif-

The characteristic slopes found for a single horizontal fracture with infinite conductivity (Gringarten
& Ramey 1974) are depending on the dimensionless
formation thickness h.Jdefined as

h.J = h ' Kr I(re Kv) [-]

Kr = conductivity of the fracture [lit]
w - fracture'swidth[I]
Km - conductivity of the matrix [I/t]
Xr .. holf-Iength of the frneture [I]


h = reservoir thickness [I]
rr ...fractureradius[I]
Kr .. radial Crncture
conductivity [1/1]
Ky '" wrliclIl f(mnnlionconductivity [1/11
Cll1co-LtlY& Snll1nnlcgo(l97K) dol1H1nMlrnlcd
that for nil practical purposesa single vertical frae.
When hi.< 1, the following characteristicnow
ture with Cr ~ 100could be regardedas an infinite phasesareobservedin a log-logplot:
conductive fracture that coincides with the GringarVery early time data shows a slope of 1 that inten et nl. (\ 974) Infinite nux model fOf Infinite con.
dudlvl! tfnctures.
Ent'ly time lIntn showstI slope ot' 0,5 tOr IIn@l\t
IfCf < \00. then the fracture has finite conductivnow
\~. 1\\,,-'t""\.~ \\\t'lM~ thnt ro"!1.\t'~~ tht t~\\\~


dUtti\'ity "ms\ be applied (Cinco.Ley&. Snmnniego 1978,Boehmer& Boonstra1986),


h\tt'n\\C\\i:\tt'ti\\\~ dnh\ ~h\,,\~ :\ ~\\'t'Cof \ ~I\\I\

tbr tntl\sict\tnow innuct\ccJby the limitedt~
ture reservoir

The flow phases for h.t~ 1, when represented in a

log-log plot. are reduced to a slope of 0.5 from very
early time data to intermediate time data, indicating
linear flow.
However, due to the very short early time period,
this phase might be masked by wellbore storage or
simplY be missed. Va1k6& Economides (1997) investigated the behavior of a horizontal penny-shaped
fracture with finite conductivity, but did not find
characteristic slopes over the dimensionless time t.J
period in the range ofO.! to 100.
In all these cases the radial-acting flow phase
commences approximately at the dimensionless time
t.J==15, which is defined as:




t =Real time [t]
T =Formation transmissivity [l2/t]
S =Formation storage coefficient [-]
Only after the radial~acting flow phase is reached
can the common analysis methods for primary aquifers be applied to determine the transmissivity T,
e.g. Jacob or Thcis. The storage coefficient S can be
estimated with equation (3), if the fracture halflength Xc is known. Figure 3 illustrates the extremely large error made in the calculation of S
when the Jacob straight-line method for time drawdown data is applied to.the data measured in an observation well located near an infinite conductive
vertical fracture with uniform flux. The straight-line
method is only applicable when the distance of the
observation well is at least 5 times that of the fracture half-length (which is in fact the REV of such a

However, Figure 3 can also be used to either determine the correct storage coefficient if the relative
position of the observation well to the fracture is
known or to determine the fracture half-length if the
storage coefficient is known.
The influence of the fracture storage can be de. scribedwith the

COf= Sf' w/(S' xc) [-]

where Sc



= Fracture storage coefficient [-].

. curves show

a transition zone characterizedby an

increasing slope from 0.5 to I with increasing COf

as a result of the leakage from the formation that
gradually replaces the storage in the fl!icture. Once
the radial-acting flow phase is.reached, all water is
provided by the formation.
. ,..





LM a Inoor fonnllion ""'" IIopo 0.5

b<uIdaty, IIopo 1





Figure 4 shows the drawdown in a vertical fracture with infmite conductivi~ and infinite flux with
COf between 10-0and IQ-IO.All curves commence
with linear flow (slope 0.5). All curves with
COf < 10-4describe the drawdown in high storage
capacity fractures and their behavior is' significantly
different to that of well bore storage. In these curves
the slope of 0.5 indicates linear fracture flow and not
linear formation flow and indicates that all discharged water is provided solely by the fracture. The



relativestoragecapacityCOf after

Ramey & Gringarten (1976) defined as:







.. ......

Id 1.1

Figure 4. Drawdown in an infinite conductive vertical fracture

(Cr - 10000) with various relative fracture capacity CDf (dimensionless dJ-awdown pd
11 T
5/(2 . Q) [-] where: s
drawdown,Q discharge rate, T - formation transmissivity)

- . .




dllWlCl If'"

Figure 3. Deviation from the real storage coefficient calculated

using the lacob's
method for data of the radialacting flow phase in obserVation wells in the vicinity of a single vertical infmite conductive
fracture with uniform flux.

One must bear in mind that the determination of

the storage coefficient S is a function of its location
and not the extraction time although the radial-acting
flow phase in the observa~on well has been reached!

The drawdown curve for COf = 10-0in Figure 4

coincides exactly with the drawdown curve observed
in a isolated single fracture for the early time data
and partly with the transient phase. It confums the
fact that the transition zone is caused by the limited
extent of the fmite fracture because the water
pumped from such a isolated single fracture is provided only by the fracture storage. This observation
leads to the interpretation that these effects i.e.,
wellbore storage, fracture storage, and the closed
boundary reservoir can be described by the same
principle of pumping from a limited reservoir.


However, the pseudo-skin effect can be used to

determine whether a well is located in a fracture
zone or not, due to the fact that, in principle, no
negative skin factor or enlarged effective radius can
be observed in a primary or continuous fractured aqui(er. One useful application is for example to distinguish between the drawdown of a horizontal fracture with infinite conductivity and the drawdown of
a well in a homogenous aquifer with wellborestorage. Unfortunately this holds true only for h.i:S 1
where the skin-factor is always negative, because the
skin-factor increases towards positive values with
h.i> I. Similarly, the method can be used for vertical

The following methods and tools can be applied for

diagnose of pumping test data in fractured aqui-

. the


Comparison of drawdown and recovery data

Pseudo-skin analysis
Straight-line analysis in log-log and semilog
Special plots
First and second derivative of the drawdown data

3.1 Comparison of drawdown and recovery phase

The fIrSt check during a pumping test data evaluation should be the comparison of the drawdown and
recovery behavior of the data if both are available.
Due to the superimposing theory, both should show
the same principal curve shapes. If a flattening is
observed in the drawdown curve, the same typical
flattening must appear in the recovery data. If this is
not the case, the flattening in the drawdown was
caused by:
Variation of the discharge rate during the drawdown phase
Discharge from a closed reservoir (limited reser-


In the first case, only the recovery data after a

time correction (Birsoy & Summers 1980) can be
used. In the second case, the late time recovery data
will show a horizontal flattening instead of a steep
increaseas it is observedduringthe latetime dataof
the drawdown phase (Streltsowa 1988). This flattening of the recovery data should never be confused
with a leaky boundary. A leaky boundary would lead
to a fully recovery, which is never the c~e in a limited reservoir.

As already discussed most of the flow phases that

appear during a pumping test in a fractured aquifer
show characteristic straight-lines either in a log-log
or in a semilog plot. However, the log-log plot provides additional information as follows:
Well bore storage shows a slope of 1
Fracture storage shows a slope between 0.5 and
1 for linear flow and 0.25 to 1 for bilinear flow.
Two parallel closed boundaries show a slope 0.5.
0 (Ehlig-Economides & Economides 1985)
Three equidistant closed boundaries U-shaped
show a slope 0.5
Limited reservoir (four closed boundaries) shows
a slope of 1.
The semilog plot provides additional information
regarding the presence of negative boundaries: .
One closed boundary doubles the slope of the
radial-acting flow straight-line

Two perpendicularsclosedboundaryquadruples
the slopeof the radial-actingflowstraight-line. .

Only one straight-line will be observed when all

the boundaries .are located equidistant to the pumping well. If this is not the case, each boundary will
increase the slope of the previous straight-line

3.2 Pseudo-skin anf!lysis

A well located within or in the vicinity of a conduit
(if the distance is below the REV) shows less drawdown than expected for wells in a homogeneous
formation. This effect is known as pseudo-skin
(Gringarten & Ramey 1974). The application of the
skin determination methods as in wells situated in
homogenous aquifers would lead to a negative skinfactor sf [-]. The skin factor sf and the drilled radius
rw are related to the effective radius roffas follows
(Sabet 1991):
roff= rw e-sf [I]

3.3 Straight-line analysis

3.4 Special plots

Besides ~e semilog and log-log plots, the following
three additional plots are very useful for the diagnose of pumping test data (Cinco-Ley & Samaniego
Linear drawdown ve~uS square root of time
Linear drawdown ve~us fourth square root of
Linear drawdown versus one divided by square
root oftime
The first plot is useful for the determination of the
linear flow behavior. The drawdown data of linear
and bilinear phases will plot on a straight line that
commences at the origin of the diagram. In the second plot the drawdown data of the bilinear flow


The equation shows that a negative skin-factor

(well in a fracture) would lead to an effective radius
of the well larger than the real radius. In the case of
a positive skin-factor (well in homogenous aquifer)
the well is effectively acting with a radius smaller
than the actual one.


phase will plot on a straight line starting at the origin. The drawdown data of a spherical flow will plot
on a straight line that commences in the origin in the


4.1 Field example 1

Two boreholes located 133.5 m apart were sited

on a 15 km long, sub-vertical (77S) fault zone
crossing the Fish River in the southern part of Namibia, which might be a potential recharge source.
The fault partly separates two low yielding formations composed at horizontal intercalated layers of
claystone, siltstone and sandstone. Both boreholes
intersect the fault at 27m below the surface. The water level in both boteholes rose immediately after the
fault was struck at a level of 906.1 m amsl (8.3 m
below surface in BHl and 5.3 m below surface in
BH2). The airlift yield was estimated at more than
100 m3fh in each borehole. Screens with 0.5 mm
slots were installed to avoid borehole collapse. Figure 5 shows the drawdown measured during one of
the constant discharge tests. Only the drawdown in
the. observation well shows a slope of 0.5 indicating
linear formation flow. However, the drawdown in
the pumped well starts almost horizontal and developes at late time to radial-acting flow. This behavior
is typical for a skin that is located at the well.

Cinco-Ley & Samaniego (l981b) demonstrated

that the first two diagrams are very useful to determine skin effects in drawdown data of wells in single fractures with either linear or bilinear drawdown
behavior. Such skin effects cause an additional
drawdown that increases clogging phenomena and in
extreme cases can even destroy the stimulation effect of drilling. in a fracture zone (Economides &
Nolte 1989).
Cinco-Ley & Samaniego (1981b) found that, due
to skin effects, the early time linear flow data in a
log-log plot are represented as an almost horizontal
Onethat develops into the radial-acting phase (bilincar flow would plot similarly). Plotting the same
data in a linear drawdown versus square root of time
diagram will also show a straight-line, but will be
shifted downwards from the origin. Both authors
stated that, by data from the pumped well only, it is
not possible to. obtain a Unique solution for the determination of the skin location, which could be 10c:atedat the well, or between fracture and formation,
or at both. However, Bardenhagen (1999) showed an
unique evaluation method for the skin location for a
single vertical fault with a relative conductivity
Cr ~ 100 by using the plot linear drawdown versus
square root of time for drawdown data of a pumped
well and an observation well located in the same
All these plots can also be applied for horizontal
features as they are only related to the flow regime.




3.5 Derivatives







The additional use of curve derivatives is of great

8dvantage because they are sensitive to small
changes in the rlrawdown curve and are independent
of skin effects, as illustrated in Figure 6. Usually, the
derivative is plotted as (tW6.t . t), which provides
the following advantages:

All characteristic


Figure 5. Example for a restticted drawdown

in a pumped well.
The slope of.0.5 in the drawdown
data of the observation
indicates linear formation flow. Simulation
for a vertical infi-

nite conductive fracture with uniform flux results: T = 0.0007; xf=.460 m; sf= 1.78.

200 m2/d; S

slopes remain the

4.2 Field example 2

A borehole was drilled into the dolomites of the
Tsumeb area, northern Naplibia. The constant discharge test indicated a short period of linear flux to a
fracture in the early stages. Water strikes were recorded at depths between 14 m and 27 m below the
water table. As soon as the water level in the
pumped well fell to the level of the rust water strike
a significant increase in the drawdown was measured, with a further increase as the water level fell
below to the second strike (Fig. 6). This behavior is
a clear indication of over-abstracting at a rate of
20 m3fh.However, due to unknown reasons an additional drawdown of 7 m can be determined from the
special plot.

Radial-acting flow phase is plotted as a horizontal fuie, which eases the identification for the
human eye
Unfortunately, derivatives applied to real data ofreo show too much noise. Smoothing of the derivatives would overcome this problem, but it cannot be
ensured that the applied mathematical algorithm
would not produce misleading artifacts. However,
recovery data are usually less noisy as they are not
iDfluenced by variations in the discharge rate and
can be used. Nevertheless, with some experience
even noisy derivative curves can be interpreted.


Cinco-Ley, H. & Samaniego, V. 1978. Transient Pressure Be,

havior for a Wellwith a Finite-ConductivityVertical Frac-

ture. SPE 6014, Richardson, Texas: Society of Petroleum

Engineers, American Institute of Mining, Metallurgical and
Petroleum Engineers.
Cinco-Ley, H. & Samaniego, V. 1981a. Transient Pressure
Analysis for Fractured Wells. SPE 7490, Richardson,
Texas: Society of Petroleum Engineers, American Institute
pfMining, Metallurgical and Petroleum Engineers.
Cinca-Ley, H. & Samaniego, V. 1981b. Transient Pressure
Analysis: Finite Fracture versus Damaged Fracture 'case.
SPE 10179, Richardson, Texas: Society of Petroleum Engineers, American Institute of Mining, Meta1\urgical and Pe-




troleum Engineers.

Figure 6. Graphical skin evaluation

of draw down curve.


Economides, MJ. & Nolte, K. G. 1989. Reservoir Stimulation.

2nded. Englewood Cliffs. New Iersey: Prentice Hall
Ehlig-Economides, C. & Economides, MJ. 1985. Pressure
Transient Analysis in an Elongated Linear Flow System.
SPE 12520, Richardson, Texas: Society of Petroleum Engineers, American Institute of Mining, Metallurgical and P.etroleum Engineers.
Gringarten, A.C., Ramey Jr., HJ., & Raghavan, R. 1974. Unsteady-state pressure distribution created by a well with a
single infinite-conductivity vertical fTacture. SPE 4051, In
Proceedings of SPE-AIME 47th Annual Fall Meeting, San
Antonio, Texas, Oct. 8-11, 1972. Richardson, Texas: Society of Petroleum Engineers, American Institute of Mining,
Meta1\urgical and Petroleum Engineers.
Gringarten, A.C., & Ramey Ir., H.I. 1974. Unsteady-state Pressure Distribution Created by a We1\ With a Single Horizontal Fracture, Partial Penetration, or Restricted Entry; Society of Petroleum Engineers Iournal, Transaction Vol. 257:
Kaiemi, H. 1969. Pressure transient analysis ofnatura1\y fTaetured reservoirs with uniform fTacture distribution. Trans.
Soc. Pet. Eng. American Institute of Mining, Metallurgical
and Petroleum Engineers AIME, 246: 451-462.
Moench, A.F. 1984. Double-Porosity Models for Fissured
Groundwater Reservoir with Fracture Skin. Water Resources Research, Vol. 20, No 7: 831-846. Ramey Ir., H.J. & Gringarten, A.C. 1976. Effect of highvolume vertical fTactures on geothermal steam well behavior. In Proceedings Second UN. Devel. And Use of Geathermal Resources, San Fransisco, U.S. Governmerit PrintingOffice, WashingtonD.C., Vol.'3: 1759-1762.
Sabet, M.A: 1991. Well Test Analysis. Contributions in Petroleum Geology and Engineering, Gulf Publishing Company,
Houston, Texas.
Streltsowa, T.D. 1988. Well Testing in Heterogeneous Formations. An Exxon Monography. New York: John Wiley &
Sons, Inc.

using the linear flow period

It is demonstrated that pumping tests can provide
various types of useful information regarding the
reservoir behavior of fractured rock aquifers, if reliable data and principal structural geological
information are available. Furthermore, it is shown
that only a detailed data analysis can produce
reliable results, using a combination of diagnostic
methods and tools. Some of the most important
functions are assuring data quality by comparing
drawdownand recoverydata.
Use pseqqo-skin analysis to overcome masking
or coinciilentaleffects
Identify different flow regimes with detailed
straight-line diagnosis Obtain skin effects with application of special
Use derivatives from real data. This requires
much experience. Nevertheless, they are especially useful to identify radial-acting flow, because they are not affected by skin effects.
However, further investigations are necessary,
especially for the Cjvaluationof data from pumping
tests performed in discontinuous aquifers below the
representative elementary volume (REV).

Va1k6, P. & Economides,

M.I. 1997. Transient Behavior ofFinite Conductivity
Horizontal Fractures. SPE 38436. Society
of Petroleum Engineers,
SPE Iournal, Vol. 2: 213-222.
Warren, I.E. & Root, P.J. 1963. The behavior of naturally fractured reservoirs. Trans. Soc. Pet. Eng. American Institute of
and Petroleum

Bardenhagen, J. 1999. Skin Localization at Wells Drilled in a
Vertical Fault Zone. GROUND WATER, Vol. 37, No 5:
Barenblatt, G.I., Zheltov, Yu.P., & Kochina, I.N. 1960. Basic
concepts in the theory of seepage of homogeneous liquids
in fISsured rocks. In Tatiana D. Streltsova (eds), Well TestIng In Heterogeneous Formations:, 370-373. An EJexon
Monography. 1988. New York: John Wi/ey & Sons, Inc.
Birsoy, Y.K. & Summers, W.K. 1980. Detennination ofaquifer parameters from step tests and intermittent pumping
data. GROUND WATER. Vol. 18: 137-146
Boehmer, W.K. & Boonstra, J. 1986. Flow to wells in intrusive
dikes. Ph.D. Thesis, Free University, Amsterdam: 262pp.

228, 3 (3): 245-255.

Wei, L., Hadwin, I., Chaput, E.,. & Swaby, P. 1998. Discriminating J;racture Patterns in Fractured Reservoirs by Pressure
Transient Tests. Prepared for presentation
at the annual SPE
Annual Technical Conference
and exhibition
held in New
Orleans, Louisana,
ciety of Petroleum


27-30 September
1998. SPE 49233.
Engineers, Inc.: 609-618.