 2.2 J.R. Fanchi, M.J. Cooksey, K.M. Lehman, A.Smith, A.C.Fanchi, C.

J Fanchi
Probabilistic Decline Curve Analysis of Barnett, Fayetteville, Haynessville, and Woodford
Gas Shales
This paper presents a probabilistic decline curve workflow to model shale gas production from the
Barnett, Fayetteville, Haynesville, and Woodford shales. Ranges of model input parameters for four gas
shales are provided to guide the preparation of uniform and triangle probability distributions. The
input parameter ranges represent realistic distributions of model parameters for specific gas shales.
A probabilistic work flow is presented that is a modification of reservoir simulation workflows and is
designed for use with rate-time decline curve models. Different workflows have been developed
previously for performing reservoir simulation. The workflows can integrate uncertainty in the
development of recovery forecasts. The workflow is presented which uses the probabilistic decline
curve analysis (DCA) workflow to develop model input parameter ranges for the Barnett, Fayetteville,
Haynesville, and Woodford shales. The methodology is inbuilt in a software program that calculates that
distribution of Estimated Ultimate Recovery (EUR) for gas production from unconventional shale gas.
Using well data from the four shale plays mentioned above a realistic minimum and maximum values
for the uniform and triangular distribution of a parameter were determined. The realistic range of
model input parameters can be used to guide the preparation of uniform and triangle probability
distributions.
Decline curve models for unconventional resources
Three decline curve models are used in this workflow which are:
 Stretched Exponential Decline Model (SEDM) which is a generalization of the Arps exponential
model. The SEDM has three parameters qi, n and τ (or a, b and c):

[ ( )] [ ( )]
n c
t t
q=q i exp − = aexp −
τ b

 Natural Logarithmic Decline Model (LNDM) is based on the logarithmic relationship between
pressure and time in adradial flow system. It has two parameters a and b.
q=alnt +b
 Arps hyperbolic decline curve model (HYDM) with the limitation of 0<b<1. It has three
parameters:
−1
q=a ( 1+bct ) b
Cumulative gas production Q is given by the integral
T
Q=∫ qdt
T0
The lower limit to is the initial time and the upper limit T is the time when the economic limit is
reached.

Probabilistic DCA workflow

Reserves estimates may be either deterministic or probabilistic. A probabilistic estimate of reserves
uses geological, engineering, and economic data to generate a range of estimates and their associated
probabilities. Linear regression can be used to obtain a deterministic estimate of reserves. The
probabilistic estimate of reserves is obtained here using the workflow for probabilistic DCA of
unconventional gas production outlined in Fig. 1

Gather rate-time data Specify input parameter Specify

distributions constraints

Uniform Constraint Options

Objective Function
Gas Rate
Triangle
Cum Gas

Apply Constraints Generate Range of

(Select Subset ) Decline Curves

Generate EUR Determine

Fig.1. Probabilistic DCA workflow.

Six steps are involved in the probabilistic decline curve analysis method. These are:
 Step DCA1: Gather rate-time data – acquire production rate as a function of time.
 Step DCA2: Select a DCA model and specify input parameter distributions – the number of
input parameters depends on the DCA chosen.
 Step DCA3: Specify constraints - available rate-time production history is used to decide which
DCA trials are acceptable.
 Step DCA4: Generate decline curve trials – decline curve trials are obtained by running the DCA
model.
 Step DCA5: Determine subset of acceptable trials – the trials generated in Step DCA4 are
compared to the criteria specified in Step DCA3.
 Step DCA6: Generate distribution of performance results – the distribution of EUR values for
the subset of acceptable trials is analyzed and the 10 th(PC10), 50TH (PC50), and 90th (PC90)
percentiles are determined.

Model results
The workflow in Fig.1 is applied using rate-time data for gas production for a sampling of wells from
the Barnett, Fayetteville, Haynesville and Woodford shales. Three DCA models (SEDM, LNDM and
HYDM) are used to match rate-time data for each well. A match of cumulative gas production at the
end of the historical production period is used as the constraint for selecting the subset of trials. The
constraint requires that the decline curve model and its set of model parameters must generate a
cumulative gas production that matches actual cumulative gas production to within 1% by the end
of the historical production period. Figs. 2-5 illustrate the quality of the matches for each shale gas
play. Each figure shows the match of the 50 th percentile (PC50) trial to well data. The parameters for
the PC50 trial are collected and a minimum and maximum value for each model parameter is
determined for the set of wells in each play.

Fig. 2. Barnett shale A-PC50 Fig. 3. Fayetteville shale A-PC50

Fig. 4. Haynessville shale A-PC50 Fig. 5. Woodford shale A-PC50

 Table 2
Barnett Shale Gas Wells.
Case Parameter SEDM LNDM HYDM

MIN MAX MIN MAX MIN MAX

PC50 A 973,170 1,014,700 239,570

43,398 5976 31,002 0.869
PC50 B 5.3E07 0.1821 0.9736
34,214 0 223,090 0 0.0212
PC50 C 0.08147 0.22797 0.244
311

Table 3
Fayetteville Shale Gas Wells.
Case Parameter SEDM LNDM HYDM

MIN MAX MIN MAX MIN MAX

PC50 A 951660 1,037,500 53,075 95,847 22,994 178,610 122,020 364,690

PC50 B 0.00372 0.39793 0.7372 0.9294
PC50 C 0.12784 0.28141 0 0 0.1011 0.3803

Table 4
Haynesville Shale Gas
Wells.
Case Parameter SEDM LNDM HYDM

MIN MAX MIN MAX MIN MAX

PC50 A PC50 B 975,540 1,034,800 91,907 4235.4 33,319 523,580

PC50 C 4.6E12 1.1276 0.65501 0.96419
34,839 0 339,460 0
0.3553 0.01153 0.52415
0.04457

Table 5
Woodford Shale Gas
Wells.
Case Parameter SEDM LNDM HYDM

MIN MAX MIN MAX MIN MAX

PC50 A 961,150 1,046,400 215,090 1979 10,518 966,990

PC50 B 6.63E20 32.7740 1,120,800 0 0.60790 1.42630
10,859 0
PC50 C 0.7969 0.02601 0.72473
0.0282
Conclusion
Shale gas production data from the four shale plays have been analyzed using probabilistic
decline curve workflow. Minimum and Maximum values of model input parameters were
determined from the model input parameters for the 50th percentile trials. The model input
parameters are realistic distribution of model parameters that can be used to guide the
preparation of uniform and triangle probability distributions.