SPE 116575

The Determination of Minimum Tested Volume from the Deconvolution of

Well Test Pressure Transients
Tim Whittle, BG Group, and Alain Gringarten, Imperial College

Copyright 2008, Society of Petroleum Engineers

This paper was prepared for presentation at the 2008 SPE Annual Technical Conference and Exhibition held in Denver, Colorado, USA, 2124 September 2008.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
Radius of investigation and tested volume are important results of well test analysis which can make or break future field
development. Currently, their evaluation is either approximate or relies on a complete analysis of the transient pressure
response using an appropriate model. A new method is proposed that uses the deconvolved derivative response to determine a
minimum tested volume. It is accurate and does not rely on further transient analysis.
The method can be applied to any oil or gas well test. It is simple and only requires input of data that is known at the time
of testing. Furthermore, if the uncertainty in the deconvolution is quantified, then the uncertainty in tested volume is also
defined.
Field examples of both oil and gas well tests are presented which demonstrate how tested volume is easily calculated.
Radius of investigation is calculated from the tested volume by making assumptions about the reservoir geometry. The method
relies on a good deconvolution algorithm which can also compute the error bars in the derivative response.
Recent advances in deconvolution algorithms have enabled the use of this simple but powerful new method to accurately
calculate tested volume without the need for complex transient analysis.
Introduction
In the last thirty years, there have been many advances in well test analysis (Earlougher1, 1977, Gringarten2, 2007) of which
deconvolution is probably the most recent. In the context of the analysis of pressure transients recorded during the testing or
production of a well, a reliable deconvolution algorithm (von Schroeter3 et al, 2001) will extract an equivalent single rate
pressure response from the actual well response which is usually a continuous variation in rate and pressure. Within its known
limitations (Levitan4, 2004), it is a powerful technique because it allows the engineer to see the transient response of the well
and reservoir without the complexity of the effects of superposition. Furthermore, the deconvolved response applies to the
entire duration of the test rather than the duration of any particular constant rate period within the test (usually a build-up).
A well test analysis work flow starts with the diagnosis of the transient response using a log-log plot of the pressure change
and derivative response. Flow regimes are identified and an appropriate model selected. Often the models are analytical and
they describe the pressure response of the well producing at constant rate for a given set of model parameters. The principle of
superposition is then used to convert the constant rate response into the response due to the actual rates at which the well
flowed. The resulting pressure history simulation is then compared with the observed data on a variety of plots. The
parameters of the model are adjusted manually or automatically using regression techniques (Levenberg-Marquardt5,6) until
the best fit between the model simulation and the observed pressure response is obtained. If a numerical model is selected, a
similar process is used but without the constraints of linearity that are imposed by the use of analytical models and
superposition (Houze7,8 et al, 2002, 2007).
Much effort has been put into developing complex models (Bourdet9 2002), to match the variety of derivative responses
that are observed during both well tests and also - with the increased use of permanently installed down hole pressure gauges during well production. There exists a multitude of models and associated parameters that will match a given observed
response and this inherent non-uniqueness in the analysis process means that a model needs to be selected carefully - all
models are wrong but some are useful (Box10, 1987). Deconvolution helps the analysis by identifying the character of the
transient response without the additional complication caused by the superposition of multiple flow rates. However, depending
on the quality of data, it may not be possible to extract a unique deconvolution. This compounds the problem of non-

SPE 116575

uniqueness. Not only is there a multitude of possible models for any one response, there may also be a multitude of responses.
This is not a limitation of deconvolution per se; rather, deconvolution highlights the uncertainty that may already be inherent
in the observed data. However, regardless of this uncertainty, one flow regime that can be readily defined is pseudo-steady
state. This is because it will always be the last flow regime (in the same way that wellbore storage is always the first).
A well flowing in a finite reservoir will eventually reach pseudo-steady state and the pressure will respond according to
material balance rules. The time to reach pseudo-steady state depends on the diffusivity of the reservoir, the distance from the
well to the furthest boundary and the influence of other producing wells. Material balance calculations (Mathews et al11, 1954,
Jones12, 1956, Dietz13, 1965, Ramey and Cobb14, 1971) can provide estimates of average reservoir pressure and the volume of
fluids in the reservoir that are connected to the well. In practice, except during extended well tests (EWT) or in smaller
reservoirs, well tests are not long enough to reach pseudo-steady state so it is not possible to calculate the total connected
volume. Instead, a minimum tested volume is calculated using techniques described by various authors (Boutaud de la Combe
et al15, 2005, Levitan16 et al, 2006, Kui-Fu Du17, 2007, Larsen and Straub18, 2007). These techniques (sometimes referred to as
Shrinking Box methods) represent the current state of the art and they require the selection of a model that applies to the
entire observed transient response. The model is bounded on all sides and the distance to the outermost boundaries are reduced
until they just start to have an influence on the pressure transient response. It can then be asserted that the connected volume
must be greater than or equal to the volume calculated within the model. Whilst being theoretically sound and accurate, there
are several difficulties with this technique. First, it can take considerable time, effort and expertise to find and match an
appropriate model. Second, the calculations require knowledge of a number of well and reservoir parameters that may not be
available at the time of the test and third, it is not obvious how to quantify the uncertainty in the result. The method proposed
in this paper aims to compliment these techniques by providing an estimate of minimum connected volume using a very
simple approach without the difficulties just described.

Pressure Change and Derivative (psi)

Discussion
Figure 1 depicts an example of a log-log plot of the pressure derivative versus time for a well producing at constant rate, q.
1000
Wellbore storage

100
Range

ssure
of Pre

nt
Transie

10

Pseudo-Steady State

1
Drawdown
Build-up derivative
0.1
0.0001

0.001

0.01

0.1

10

100

1000

Elapsed time (hrs)

Figure 1: Example Pressure Transient Between Wellbore and Reservoir Storage
The transient part of the response occurs between two parallel unit slope straight lines. The first line represents well bore
storage and is controlled by the volume of the well completion, Vc and the compressibility of the fluid in it, cf:

p =

qBt
24c f Vc

Equation 1

The second line represents pseudo-steady state and is controlled by the reservoir pore volume, Vr, and its total compressibility,
ct:

p =

qBt
24c tVr

Equation 2

SPE 116575

Total compressibility is defined:

Equation 3

ct = cr + S o co + S g cg + S wcw

The shift in log time between the first and second line is directly proportional to the ratio of the storage of the reservoir, Vrct,
and the storage of the well, Vccf. The entirety of the drawdown pressure transient response lies in between these two extreme
unit slope straight lines, i.e. prior to storage and after pseudo-steady state, there are no transients that can be used for well and
reservoir characterization. Normally, a test is not long enough to reach the second line and, furthermore, the derivative
response of a build-up will deviate from the drawdown. As pseudo-steady state is approached, it will display a sharp
downward trend rather than the unit slope which is only characteristic of pseudo-steady state for a constant rate drawdown.
Deconvolution reconstructs the drawdown response such that, if the test is long enough and pseudo-steady state is reached,
a unit slope will display in late time and the reservoir pore volume, Vr, can be calculated from any point on that unit slope
straight line:

Vr =

qBt pss
24ct ppss

Equation 4

If the test is not long enough to reach pseudo-steady state, then the proposed approach is to fit a unit slope straight line
through the last point of the deconvolved derivative response (Figure 2). Such a line defines the earliest time that pseudosteady state could occur without it being evident in the actual test response and therefore represents the smallest reservoir that
could be connected to the well. This is the definition of minimum connected volume and it is simply calculated using Equation
4. If it is assumed that the well produces single phase fluid and that the reservoir contains either oil or gas with irreducible
water saturation, Sw, then either the minimum tested stock tank oil initially in place (STOIP) or the minimum tested gas
initially in place (GIIP) is calculated using Equation 5 and Equation 6 respectively.

STOIIPmin tested =

(1 S w )qtmax

Equation 5

24ct pmax

GIIPmin tested =

(1 S w )qtmax

Equation 6

24ct nm( p )max

Pressure Change and Derivative (psi)

1000

Unit
Slope
(pss)

Deconvolved Derivative

pmax

100

Build-up Derivative

10

1
0.001

0.01

0.1

1
Elapsed Time (hrs)

10

100

t max 1000

Figure 2: Selecting the last point of deconvolution derivative to define pseudo-steady state

For gas reservoirs, see Appendix 1 for additional definitions.

SPE 116575

Examples
Example 1 is a gas well test. The well produced over a five day period at a maximum rate of 46 MMscf/d. At the end of the
main flow period, the well was producing at a rate of 40.4 MMscf/day. It was shut-in for a 29 hour build-up. The calculation of
minimum tested gas volume based on the deconvolution of the main build-up is summarized in Figure 3. The total volume of
gas produced during the test was 57 MMscf which is 1.6% of the calculated minimum tested volume of 3.87 bscf. A full
analysis of the data using an analytical closed rectangular reservoir model and the Shrinking Box method yields a match as
shown in Figure 4 with an initial gas in place of 3.91 bscf.

100

8200

90
Pressure

80

Pressure (psia)

7800

70

7600

60
50

7400
Gas Rate

7200

40
30

7000

20

6800
30

40

Input:
Sw=0.15
ct = 6.62E-5 1/psi
q = 40.4 MMscf/d

p=8135.32 psia
= 0.032 cp
z = 1.247

50

Pseudo-Pressure and Derivative (psi**2/cp

6600
20

10
60

Gas Rate (MMscf/D)

8000

GIIPmin tested =

GIIPmin tested =

(1 S w ) q
24ct

t max 2 p
m( p ) max z

(1 0.15) 40.4 93.6 2 8135.32

24 6.62 E 5
2.3 E 8 0.0032 1.247

GIIPmin tested = 3534MMscf = 3.53bscf

0
70 80 90 100 110 120 130 140
Elapsed Time (hrs)

1E+9

Unit Slope
(pss)

Deconvolved
Response

1E+8

Build-up
Response

1E+7

t max = 93.6hrs
m( p )max = 2.30 E8 psi 2 / cp
1E+6
0.0001

0.001

0.01

0.1
1
Elapsed Time (hrs)

10

100

1000

Figure 3: Example 1. Gas test

8200

Observed Data
Model Simulation
Modelled GIIP: 3.91 bscf

Pressure (psia)

8100
8000
7900
7800
7700
7600
7500
7400
40

60

80

100

120

140

Elapsed Time (hrs)

Figure 4: Example 1: Simulation using a closed reservoir model and the Shrinking Box method

SPE 116575

In Example 1, the deconvolution method agrees very closely with the shrinking box method. In other data sets where the
production time is short compared to the shut-in time, it has been noticed that the deconvolution method under predicts
compared to the shrinking box. One possible reason for this is that the placing of the pseudo steady state unit slope line
directly on the last point of the derivative implies a sharp change in slope (i.e. a discontinuity in the the 2nd derivative). In
reality, diffusion creates a smooth transition and therefore the pseudo-steady state unit slope is likely to start later in time. This
is discussed further in Appendix 2.
Example 2 is a post-acid well test from an oil reservoir. Pre- and post-acid production is available but the bottom hole
pressure data is limited to the post-acid response. In such a case, there exists significant uncertainty in the initial reservoir
pressure and this is reflected in the uncertainty in the deconvolution which, in turn, impacts the uncertainty in the calculation
of minimum tested volume (Figure 5). If it is assumed that the deconvolved derivative represents P50 case and that the error
bars represent the P10 and P90 cases, then it can be stated that there is a 90 percent chance that the test has seen a volume
greater than 61.7 MMstb and a ten percent chance that it has seen a volume greater than 320 MMstb. It is important to realize
that the actual reservoir could be substantially larger in size since these volumes are only those seen by the test.
8000
Pressure

7800

6000
5000

7600

4000
7400

3000

7200

2000

Oil Rate

STOIPmin tested =

7000
Oil Rate (STB/D)

Pressure (psia)

8000

100

200
300
Elapsed time (hrs)

pmax

24ct

P90

P50

Sw

1000
7000

(1 S w ) q tmax

0
400

P10

0.129

Stb/day

ct

1/psi

2380
9.44e-6

hrs

tmax

304

psi

pmax
STOIPtested

MMstb

45

28

8.7

61.7

99.3

320

Pressure Change and Derivative (psi)

1000

Pressure Change
Unit Slope
(pss)

Build-up Response
Derivative

100

= 8.7, 28, 45 psi

pmax
Deconvolved
Derivative with error bars

10

1
0.001

0.01

0.1

1
Elapsed Time (hrs)

10

( P10, P50, P90)

100

1000

t max = 304hrs

Figure 5: Example 2. Oil test with uncertainty in deconvolution

Radius of Investigation
In order to calculate a radius of investigation, a model has to be assumed. For example, if it assumed that the reservoir is single
layer, uniformly thick, infinite with constant permeability and porosity, then a good estimate of radius of investigation1 is:

ri = 0.029

kt
ct

Equation 7

An alternative radius of investigation can be defined using the intersection of the zero slope derivative response due to infinite
acting radial flow and the unit slope corresponding to pseudo-steady state (Figure 6):

pD int = 0.5 = 2t DA int

Equation 8

SPE 116575

Pressure & Derivative

pint
Elapsed Time

t int

Figure 6: Intersection of Infinite Acting Radial Flow and Pseudo-Steady State

Assuming a circular reservoir of area,

A = re2 , Equation 8 can be re-written in real terms:

re = 0.2729

qBtint

ct hpint

Equation 9

For any reservoir, it is proposed, to use a distance of investigation, di, in combination with a coefficient of reservoir
complexity, CRC, from the deconvolved derivative defined as follows:

CRC =

pmax

pint

Equation 10

And then:

d i = CRC re

Equation 11

For a given distance of investigation the coefficient or reservoir complexity provides the equivalent reduction in flow area as
compared to circular drainage radius. Undoubtedly the CRC is closely related to the reservoir shape factor, CA as defined by
Dietz13 and has a very important influence on the future deliverability of the reservoir.
As with the calculation of minimum tested volume, the uncertainty in the deconvolution (including the effects of gauge
resolution which can have an important influence) will carry through to the uncertainty in radius of investigation.
Conclusions
Deconvolution provides a single rate pressure transient derivative response from which minimum tested volume can be easily
calculated with minimum input data. The method applies in both liquid (oil or water) and gas reservoirs assuming single phase
flow.
The uncertainty in deconvolution, which is mainly due to unknown initial pressure, errors in rate history and finite gauge
resolution, defines and quantifies the uncertainty in the minimum tested volume.
A new method of calculating distance of investigation is proposed which is designed to be used in combination with a
coefficient of reservoir complexity also defined in the paper.
The methods do not require the use of any particular model but implicit in any deconvolution algorithm is the assumption
of single phase fluid flow (diffusion) in a porous medium.
Acknowledgements
Thanks to BG Group for permission to submit the paper and to colleagues, Dave Paul and Nigel Robertson, for ideas and
provision of example datasets.

In Equation 9, for gas, if q is in MMscf/d then B is in bbl/MMscf. i.e. the product qB is in bbl

SPE 116575

Nomenclature
q Flow rate (stb/day for oil : MMscf/day for gas)
B Formation Volume Factor (rb/stb for oil : cf/scf for gas)
t Elapsed time (hrs)
p pressure derivative (psi)
c Compressibility (1/psi)
V Volume (bbl for oil : MMscf for gas)
S Saturation
STOIP Stock Tank Oil Initially In Place (stb)
GIIP Gas Initially In Place (MMscf)
m(p) Pseudo-pressure (psi2/cp)
nm(p) Normalized pseudo-pressure (psi)
r Radius (ft)
k Permeability (mD)
Porosity
- Viscosity (cp)
CRC Coefficient of Reservoir Complexity
Subscripts:
f - Fluid
c - Completion
t - Total
r - Reservoir/rock
o - Oil
g - Gas
w - Water
pss - Pseudo-steady state
max - Maximum
min tested Minimum tested
D Dimensionless
A - Area
e - Extent
int - Intercept
i - Investigation

SPE 116575

References
1

Earlougher, R. C.: Advances in Well-test Analysis. Vol. 5, Society of Petroleum Engineers Monograph, Dallas, Texas
(1977).
2
Gringarten, A. C.: From Straight-lines to Deconvolution: the Evolution of the State-of-the-art in Well Test Analysis, paper
SPE 102079 presented at the 2006 SPE Annual Technical Conference and Exhibition, San Antonio, TX, 24-27 Sep.
3
von Shroeter, T., Hollander, F., Gringarten, A.:Deconvolution of Well Test Data as a Nonlinear Total Least Square
Problem, paper SPE 71574, presented at the 2001 SPE Annual Technical Conference and Exhibition, New Orleans, LA, 30
Sep.-3 Oct.
4
Levitan, M. M, Crawford, G. E. and Hardwick, A.: Practical considerations for Pressure-Rate Deconvolution of Well Test
Data, paper SPE 90680 presented at the 2004 SPE Annual Technical Conference and Exhibition, Houston, TX, 26-29 Sep.
5
Kenneth Levenberg (1944). A Method for the Solution of Certain Non-Linear Problems in Least Squares. The Quarterly of
Applied Mathematics 2: 164168.
6
Donald Marquardt (1963). An Algorithm for Least-Squares Estimation of Nonlinear Parameters. SIAM Journal on Applied
Mathematics 11: 431441. doi:10.1137/0111030.
7
Houze, O, P.: Why We Should Stop Using Pseudopressures and Other Good Old Well Test Interpretation Tools After So
Many Years of Good Service paper SPE 77619 presented at the 2002 SPE Annual Technical Conference and Exhibition, San
Antonio, TX, 29 Sep.2 Oct.
8
Houze, O. P., Viturat, D., Fjaere, O.,S.: Dynamic Flow Analysis, http://www.kappaeng.com/downloads/default.aspx?s=16,
Kappa Engineering, 2007.
9
Bourdet, D.: Well Test Analysis: The Use of Advanced Interpretation Models, Elsevier, (2002) 1.
10
Box, George E. P.; Draper, Norman R.: Empirical Model-Building and Response Surfaces, p. 424, Wiley, 1987. ISBN
0471810339.
11
Matthews, C.S., Brons, F. and Hazebroek, P.: A Method for Determination of Average Pressure in a Bounded Reservoir,
Trans., AIME (1954) 201, 182
12
Jones, P.: Reservoir Limit Tests, Oil and Gas J. (June 18, 1956) 54, no59, 184
13
Diez, D.N.: Determination of Average Reservoir Pressure From Build-Up Surveys, J. Pet. Tech. (Aug. 1965) 955-959;
Trans.,AIME, 234.
14
Ramey, H. J., Jr. and Cobb, W.M.: A General Pressure Build-up Theory for a Well in a Closed Drainage Area J. Pet.
Tech. (Dec., 1971) 1493-1505; Trans., AIME (1971), 252.
15
Boutaud de la Combe, J.L., Akinwumni, O., Dumay, Ch., Tachon, M., Use of DST for Effective Dynamic Appraisal : Case
Studies From Deep Offshore West Africa and Associated Methodology paper SPE 97113, Presented at the 2005 SPE Annual
Technical Conference and Exhibition, Dallas, Texas, Oct. 9-12.
16
Levitan, M.M., Ward, M.J., Boutaud de la Combe, J.-L. and Wilson, M.R., The use of Well Testing for Evaluation of
Connected Reservoir Volume paper SPE 102483 presented at the 2006 SPE Annual Technical Conference and Exhibition,
San Antonio, TX, 24-27 Sep.
17
Kui-Fu Du, Use of Advanced Pressure Transient Analysis Techniques To Improve Drainage Area Calculations and
Reservoir Characterisation: Field Case Studies paper SPE 109053 presented at Offshore Europe 2007, Aberdeen, Scotland,
U.K., 4-7 Sep.
18
Larsen, L. and Straub, R.B.: Determination of Connected Volume and Connectivity From Extended Tests in
Compartmentalised and Layered Reservoirs paper SPE 110011 presented at the 2007 SPE Annual Technical Conference and
Exhibition, Anaheim, CA, 11-14 Nov.

SPE 116575

Appendix 1
Application to gas
The techniques described in the paper apply to gas reservoirs if it is assumed that, over the period of time that the
deconvolution is calculated, the depletion is small (i.e. the average pressure, and therefore the reservoir gas compressibility,
remains fairly constant). The assumption is justified provided that, at the time when pseudo-steady state is reached, the
cumulative production of the well is a small proportion of the total reservoir volume.
The same equations apply except that the derivative, p , is replaced by the normalized pseudo- pressure derivative,
nm( p) , with normalized pseudo-pressure defined: :
p

nm( p ) =

z 2 p
dp
2 p 0 z

Equation A- 1

With average gas viscosity, , and compressibility factor, z , defined at average reservoir pressure, p , and reservoir
temperature.
If pseudo-pressure is used without normalisation then the z term has to be included explicitly as shown in Example 1.
2p

Appendix 2
Transition time to pseudo-steady state
The transition time multiplication factor from the end of one flow regime to the start of pseudo-steady state is estimated
using analytical models and shown in Figure A- 1. It is plotted versus the change in angle, , of the derivative slope (i.e. the
second derivative) which is defined:

d (log p)

d (log t ) t =tmax

= 45 tan 1

Equation A- 2

That is, a change from radial flow (derivative slope zero) to pseudo-steady state (slope 1) is 45 degrees; linear flow (slope 0.5)
to pseudo-steady state is 18.4 degrees and spherical flow (slope 0.5) to pseudo-steady state is 71.6 degrees.

Transition Time Factor

4
3.5
3

Spherical

2.5

Radial

2
Linear

1.5

y = -0.0004x + 0.0607x + 1
2
R =1

Pseudo-steady state

0.5
0
0

20

40

60

80

Transition Angle
Figure A- 1: Transition Time Factor versus Transition Angle
A polynomial fit to these points (and the limiting case where the derivative has already reached pseudo-steady state) allows an
estimate of the transition time factor for any derivative slope from -1/2 through 1. It can be argued that the minimum tested
volume calculated from the deconvolution derivative can be increased by this transition time factor.