You are on page 1of 16

Refracturing Works

Applicable in gas or oil wells, fracture restimulations bypass near-wellbore damage, reestablish good
connectivity with the reservoir and tap areas with higher pore pressure. An initial period of production
also can alter formation stresses, resulting in better vertical containment and more lateral extension
during hydraulic fracturing, and may even allow the new fracture to reorient along a different azimuth.
As a result, refracturing often restores well productivity to near original or even higher rates.

George Dozier
Houston, Texas, USA
Jack Elbel
Consultant
Dallas, Texas
Eugene Fielder
Devon Energy
Oklahoma City, Oklahoma, USA
Ren Hoover
Fort Worth, Texas
Stephen Lemp
Calgary, Alberta, Canada
Scott Reeves
Advanced Resources International
Houston, Texas
Eduard Siebrits
Sugar Land, Texas
Del Wisler
Kerr-McGee Corporation
Houston, Texas
Steve Wolhart
Pinnacle Technologies
Houston, Texas

38

The potential benefits of refracturing have


intrigued oil and gas operators for more than
50 years. Most intriguing is that, under certain
conditions, this technique restores or increases
well productivity, often yielding additional
reserves by improving hydrocarbon recovery. The
approximately 70,000 new wells that are drilled
annually represent only about 7 to 8% of the
total number of producing wells worldwide.1
Therefore, getting the most output from the
more than 830,000 previously completed wells is
essential for field development, production
enhancement and reservoir management. Even
modest production increases from a portion of
the vast number of existing wells represent significant incremental reserve volumes. Refracturing
is one means of accomplishing this objective.
More than 30% of fracturing treatments are
performed on older wells. Many are completions
of new intervals; others represent treatments on
producing zones that were not fractured initially
or a combination of new intervals and previously
understimulated or unstimulated zones. An
increasing number of jobs, however, involve
refracturing previously stimulated intervals after
an initial period of production, reservoir-pressure
drawdown and partial depletion. These types of
restimulations are effective in low-permeability,
naturally fractured, laminated and heterogeneous formations, especially gas reservoirs.

If an original fracturing treatment was inadequate or an existing proppant pack becomes


damaged or deteriorates over time, fracturing
the well again reestablishes linear ow into the
wellbore. Refracturing can generate higher conductivity propped fractures that may penetrate
deeper into a formation than the initial treatment. But not all restimulations are remedial
treatments to restore productivity; some wells
that produce at relatively high rates also may be
good candidates for refracturing. In fact, the
better wells in a field often have the highest
restimulation potential.2
Wells with an effective initial treatment also
can be retreated to create a new fracture that
propagates along a different azimuth than the
original fracture. In formations with lower
permeability in a direction perpendicular to the
original fracture, a reoriented fracture exposes
more of the higher matrix permeability. In these
cases, refracturing significantly improves well
production, and supplements infill drilling.
For this reason, operators should consider
restimulation during the field-development
planning process.
Many companies, however, are reluctant to
retreat wells that produce at reasonably economic rates. The tendency is not to refracture
any wells, or to restimulate only poorly performing wells. This lack of condence and the negative

For help in preparation of this article, thanks to Curtis Boney,


Leo Burdylo, Chris Hopkins and Lee Ramsey, Sugar Land,
Texas, USA; Phil Duda, Midland, Texas; Chad Gutor, formerly
with Enerplus, Calgary, Alberta, Canada; Stephen Holditch
and Valerie Jochen, College Station, Texas; and Jim Troyer,
Enerplus, Calgary, Canada.

CoilFRAC, DSI (Dipole Shear Sonic Imager), FMI (Fullbore


Formation MicroImager), FracCADE, InterACT, Moving
Domain, NODAL, ProCADE and StimMAP are marks of
Schlumberger.

Oileld Review

2003

1993

preconceptions about refracturing are changing


because of a better understanding of refracturing
mechanics and the favorable results reported by
companies that apply this technique regularly.
To be successful, refracturing treatments
must create a longer or more conductive
propped fracture, or expose more net pay to the
wellbore compared with existing well conditions
prior to restimulation. Accomplishing these
objectives requires knowledge of reservoir and
well conditions to understand why restimulations succeed and to improve future treatments
based on experience. Quantifying average reservoir pressure, permeability-thickness product,
and effective fracture length and conductivity
both before and after refracturing allows engineers to determine the reasons for poor well
performance before new treatments and the
causes of restimulation success or failure.

Autumn 2003

Improved diagnostic techniques, such as


short shut-in time well tests, help determine the
current stimulation condition of a well and
verify refracturing potential. Advances in
fracture modeling, design and analysis software
also have contributed signicantly to restimulation success during the past ten years, as have
better candidate selection, innovative stimulation fluids, improved proppants and proppant
owback control.
This article presents results from a twoyear refracturing study and subsequent field
trials. We also discuss reasons for restimulation
success, including candidate-selection methods
and criteria, causes of underperformance in
fracture-stimulated wells, formation-stress reorientation and treatment-design considerations.
Recent examples from the USA and Canada
demonstrate refracturing implementation and
productivity improvement.

1. International Outlook: World Trends, World Oil 224,


no. 8 (August 2003): 2325.
2. Niemeyer BL and Reinart MR: Hydraulic Fracturing of a
Moderate Permeability Reservoir, Kuparuk River Unit,
paper SPE 15507, presented at the SPE Annual Technical
Conference and Exhibition, New Orleans, Louisiana, USA,
October 58, 1986.
Pearson CM, Bond AJ, Eck ME and Lynch KW: Optimal
Fracture Stimulation of a Moderate Permeability
Reservoir, Kuparuk River Unit, Alaska, paper SPE 20707,
presented at the SPE Annual Technical Conference
and Exhibition, New Orleans, Louisiana, USA,
September 2326, 1990.
Reimers DR and Clausen RA: High-Permeability
Fracturing at Prudhoe Bay, Alaska, paper SPE 22835,
presented at the SPE Annual Technical Conference and
Exhibition, Dallas, Texas, USA, October 69, 1991.

39

reservoirs in the Michigan and Appalachian


regions as well as conventional sandstone (CS)
and conventional carbonate (CC) formations
in the San Juan basin and areas of the midcontinent and Texas.
The 1996 GTI work concluded that documented refracturing treatments had yielded
incremental reserves at about $0.10/Mcf to
$0.20/Mcf, much less than the average costs for
acquiring or for finding and developing gas
reserves of $0.54/Mcf and $0.75/Mcf, respectively. Despite the potential economic benets,
operators remained reluctant to refracture
wells. Poor candidate selections appeared to be
the main reason for lack of restimulation
success and acceptance among operators.
In response, GTI funded another project in
1998 to develop specialized restimulation technology and analysis techniques. The need for
this project was underscored by anecdotal observations from the 1996 investigation that 85% of
refracturing potential in a given field exists in
about 15% of the wells. Identifying these top
candidates is crucial to restimulation success.
However, operators often perceive comprehensive eld-wide studies to be too costly in terms
of money and manpower for companies operating unconventional reservoirs, especially when
gas prices are low.

A Multiple-Basin Evaluation
Some operators report disappointing results
when refracturing previously stimulated wells,
despite documented successes in individual
wells and several field-wide restimulation
efforts.3 However, recent research, subsequent
field trials and the ongoing refracturing
programs of a few operators still attract
considerable interest and attention within the
oil and gas industry.
In 1996, the Gas Research Institute (GRI),
now Gas Technology Institute (GTI), began
investigating fracture restimulation as a low-cost
means of enhancing gas production and adding
recoverable reserves. This preliminary evaluation identified significant onshore gas
potentialconservatively more than 10 Tcf
[286.4 billion m3] of incremental reservesin
the USA, excluding Alaska (below).
These additional gas reserves are located in
the Rocky Mountain, Midcontinent, East Texas
and South Texas regions, primary in lowpermeability, or tight-gas, sandstones (TGS)
and in other unconventional reservoirs that
include gas shales (GS) and coalbed methane
(CBM) deposits (see Producing Natural Gas
from Coal, page 8). Other areas of the USA with
refracturing potential include unconventional

Michigan

Green River

GS

Denver-Julesburg
TGS
TGS

San Juan
CS, TGS, CBM

CS
Hugoton
TGS
CC
Permian
Delaware

Conventional sands (CS)


Conventional carbonates (CC)
Tight-gas sands (TGS)
Coalbed methane (CBM)
Gas shales (GS)

CC

Appalachian

USA

Piceance
TGS

CC

Val Verde
TGS

TGS
GS

Anadarko
Barnett Shale
GS

Black Warrior
CBM

CS East Texas
TGS

South Texas
TGS

400

800

1200

250

500

750

1600 km
1000 miles

> Areas with refracturing potential in the USA. The 1996 Gas Technology Institute (GTI) restimulation
investigation evaluated a wide range of gas reservoirs, including conventional sandstone and carbonate formations, tight-gas sands, gas shales and coalbed methane deposits. This evaluation focused
on conventional gas-producing provinces with cumulative production greater than 5 Tcf [143.2 billion
m3] for further evaluation. Higher production implied high numbers of older wells and more refracturing opportunities. The study also identied tight-gas sand areas with an estimated ultimate recovery
(EUR) greater than 1 Tcf [28.6 billion m3] and the largest gas shale and coalbed methane developments, but did not include offshore developments with limited production and recovery information.

40

3. Parrot DI and Long MG: A Case History of Massive


Hydraulic Refracturing in the Tight Muddy J
Formation, paper SPE 7936, presented at the SPE
Symposium on Low-Permeability Gas Reservoirs,
Denver, Colorado, USA, May 2022, 1979.
Conway MW, McMechan DE, McGowen JM,
Brown D, Chisholm PT and Venditto JJ: Expanding
Recoverable Reserves Through Refracturing, paper
SPE 14376, presented at the SPE Annual Technical
Conference and Exhibition, Las Vegas, Nevada, USA,
September 2225, 1985.
Hunter JC: A Case History of Refracs in the Oak Hill
(Cotton Valley) Field, paper SPE 14655, presented at the
SPE East Texas Regional Meeting, Tyler, Texas, USA,
April 2122, 1986.
Olson KE: A Case Study of Hydraulically Refractured
Wells in the Devonian Formation, Crane County, Texas,
paper SPE 22834, presented at the SPE Annual Technical
Conference and Exhibition, Dallas, Texas, USA,
October 69, 1991.
Fleming ME: Successful Refracturing in the North
Westbrook Unit, paper SPE 24011, presented at the
SPE Permian Basin Oil and Gas Recovery Conference,
Midland, Texas, USA, March 1820, 1992.
Hejl KA: High-Rate Refracturing: Optimization and
Performance in a CO2 Flood, paper SPE 24346, presented
at the SPE Rocky Mountain Regional Meeting, Casper,
Wyoming, USA, May 1821, 1992.
Pospisil G, Lynch KW, Pearson CM and Rugen JA:
Results of a Large-Scale Refracture Stimulation
Program, Kuparuk River Unit, Alaska, paper SPE 24857,
presented at the SPE Annual Technical Conference and
Exhibition, Washington, DC, USA, October 47, 1992.
Hunter JL, Leonard RS, Andrus DG, Tschirhart LR and
Daigle JA: Cotton Valley Production Enhancement Team
Points Way to Full Gas Production Potential, paper
SPE 24887, presented at the SPE Annual Technical
Conference and Exhibition, Washington, DC, USA,
October 47, 1992.
Reese JL, Britt LK and Jones JR: Selecting Economic
Refracturing Candidates, paper SPE 28490, presented at
the SPE Annual Technical Conference and Exhibition,
New Orleans, Louisiana, USA, September 2528, 1994.
Fengjiang W, Yunhong D and Yong L: A Study of
Refracturing in Low Permeability Reservoirs, paper
SPE 50912, presented at the SPE International Oil &
Gas Conference and Exhibition, Beijing, China,
November 26, 1998.
4. Type curves help interpret transient-pressure buildup
tests that differ from conventional semilog, or Horner,
analysis radial-ow behavior. Type curves are groups of
paired pressure changes and their derivatives generated
from analytical solutions of the diffusion equation with
strategically dened boundary conditions. Near-well
boundary conditions include constant or variable wellbore storage, partial reservoir penetration, composite
radial damage or altered permeability, and propped
hydraulic fractures. Borehole trajectory can be vertical,
angled, or horizontal. Distant boundary conditions include
sealing or partially sealing faults, intersecting faults and
sealing or constant-pressure rectangular boundaries.
The diffusion equation can be adjusted to accommodate
reservoir heterogeneity, such as dual porosity or layering. Commercial software generates type-curve families
that account for superposition in time due to ow-rate
variations before and even during transient-pressure
data acquisition. Automated regression analysis can
match acquired data with a specic type curve.
5. Reeves SR, Hill DG, Tiner RL, Bastian PA, Conway MW
and Mohaghegh S: Restimulation of Tight Gas Sand
Wells in the Rocky Mountain Region, paper SPE 55627,
presented at the SPE Rocky Mountain Regional Meeting,
Gillette, Wyoming, USA, May 1518, 1999.
Reeves SR, Hill DG, Hopkins CW, Conway MW, Tiner RL
and Mohaghegh S: Restimulation Technology for Tight
Gas Sand Wells, paper SPE 56482, presented at the SPE
Annual Technical Conference and Exhibition, Houston,
Texas, USA, October 36, 1999.

Oileld Review

High
Type curves
Interpretation requirements

Project participants, including Advanced


Resources International, Schlumberger,
Intelligent Solutions, Ely and Associates,
Stim-Lab and Pinnacle Technologies, believed
that developing an effective methodology to
identify wells with restimulation potential was
one way to expand refracturing applications.
There were three other objectives: demonstrate
productivity enhancement and recovery
improvement from refracturing, identify reasons
for underperformance in previously fractured
wells, and evaluate new fracturing techniques
and technology.
The 1998 GTI study evaluated three methods
for identifying refracturing potential that were
then tested in different types of reservoirs.
These candidate-selection methods included
production statistics, pattern-recognition technologyspecifically neural networks, virtual
intelligence and fuzzy logicand production
type curves (right).4
All three methods were used to select restimulation candidates at eld locations with at least
200 to 300 wells.5 Three sites in the USAGreen
River basin, Wyoming, USA; East Texas basin,
Texas; and Piceance basin, Colorado, USA
were chosen and actively evaluated (below): A
fourth site in South Texas was identied, but not
pursued during the GTI project. Subsequent

e
Tim

and

t
cos

se

rea

inc

Virtual intelligence
Production statistics

Low
Low

Data requirements

High

> Candidate-selection methods. The GTI project developed a methodology for


identifying wells with restimulation potential that used production statistics,
virtual intelligence and production type curves. By design, these techniques
progressed from a simple, nonanalytical statistical approach with minimal
data requirements to detailed engineering analyses requiring increasingly
comprehensive data.

reservoir studies, however, have generated


recent refracturing activity in this area (see
Production-Enhancement Evaluation, page 52).
Of the nine wells eventually treated at the
three active project sites, eight were refracturing

treatments and one was an attempted damageremoval treatment. As the project


progressed, treatment designs trended away
from high-viscosity polymer-base systems to
fracturing fluids with lower and lower gel concentrations, or slick water. Most treatments

Green River basin GTI site

East Texas basin GTI site

Piceance basin GTI site

Operator:
Enron Oil and Gas, now EOG resources.

Operator:
Union Pacific Resources Company (UPRC), now
Anadarko Petroleum Corporation.

Operator:
Barrett Resources, now Williams Company.

Formation:
Upper Cretaceous Frontier.

Formation:
Cotton Valley.

Formation:
Mesaverde group, Upper Cretaceous Williams Fork.

Location:
Big Piney/LaBarge complex, northern Moxa Arch area,
southwestern Wyoming, USA.

Location:
Carthage Gas Unit (CGU) field near
Carthage, Panola County, Texas, USA.

Location:
Parachute and Grand Valley fields near Rulison,
Garfield County, Colorado, USA.

Deposition:
Marine sandstones, primarily rivers and streams, or
fluvial and distal shore zones.

Deposition:
Complex marine sandstones, primarily barrier reef and
tidal zone.

Deposition:
Marine sandstones, primarily fluvial and marsh,
or paludal.

Reservoir:
Tight-gas sands with permeability of 0.0005 to 0.1 mD
in up to four productive horizons, consisting of as many
as eight separate intervals, or benches.

Reservoir:
Heterogeneous, highly laminated and compartmentalized
tight-gas sands with permeability of 0.05 to 0.2 mD.

Reservoir:
Compartmentalized tight-gas sands with permeability
of 0.1 to 2 mD. Because of natural fractures, effective
permeability is 10 to 50 mD.

Initial completions:
One to three stages of a crosslinked guar fluid and
nitrogen foam with 100,000 to 500,000 lbm [45,359 to
226,796 kg] of proppant sand.

Initial completions:
Three to four stages of a crosslinked fluid and proppant
volumes of 1 to 4 million lbm [453,592 to 1,814,370 kg]
for an entire well; 1996 to present, UPR and Anadarko
used slick-water fluids with less than 250,000 lbm
[113,398 kg] of proppant.

Initial completions:
Two to five stages with proppant volumes of 50,000
to 650,000 lbm [22,680 to 294,835 kg] per stage.

GTI restimulations:
Three refracturing treatments and one gel-cleanup
treatment.

GTI restimulations:
Three refracturing treatments.

GTI restimulations:
Two refracturing treatments.

> The 1998 GTI restimulation study to evaluate refracturing candidate-selection methods at three USA test sites.

Autumn 2003

41

Site
field/basin

Well

Date

Incremental
recovery, MMcf

Treatment
cost, $

Reserve
cost, $/Mcf

Success/
failure

Big Piney
and LaBarge/
Green River

GRB 45-12
GRB 27-14
NLB 57-33
WSC 20-09

Jan. 1999
Jan. 1999
Apr. 1999
Jun. 2000

602
(186)
0
302

87,000
87,000
20,000
120,000

0.14
NA
NA
0.40

S
F
F
S

Rulison/
Piceance

Langstaff 1
RMV 55-20

Jun. 2000
Jun. 2000

282
75

50,000
70,000

0.18
0.93

S
F

Carthage/
East Texas

CGU 15-8
CGU 10-7
CGU 3-8

Nov. 1999
Jan. 2000
Jan. 2000

270
407
1100

100,000
100,000
100,000

0.37
0.25
0.09

S
S
S

2852

734,000

317

82,000

Total
Average
2864 m3/d

5727 m3/d

0.26
8590 m3/d

11,455 m3/d

450
CGU 3-8
RMV 55-20

Post-restimulation rate, Mcf/D

400

CGU 10-7

CGU 15-8

350
GRB 45-12
300
250
Langstaff 1

200

WSC 20-09
NLB 57-33

150
100
50

GRB 27-14
0
0

50

100

150

200

250

300

350

400

Pre-restimulation rate, Mcf/D

> GTI eld-test results. Two of the four wells in the Frontier formation (Green River basin), all three of
the wells in the Cotton Valley formation (East Texas basin), and one of the two wells in the Williams
Fork formation (Piceance basin) were successful. Of the three unsuccessful treatments, one added
incremental reserves at a cost of $0.93/Mcf and two had mechanical or design problems. Of the latter
two, in one, the damage-removal treatment could not be pumped at the injection rate required to uidize the original proppant pack and remove suspected residual gel damage from the initial treatment;
the other failed to clean up because energized uids were not used as recommended in the GTI design.

included nitrogen [N2] or carbon dioxide [CO2]


to assist in post-stimulation cleanup, singlestage pumping schedules and ball sealers for
fluid diversion to reduce cost compared with
multistage treatments.
Standard decline-curve analysis determined
estimated ultimate recovery (EUR) for each
well; estimated treatment cost provided an
undiscounted cost of incremental reserve
additions. Costs for diagnostic tests conducted
for research purposes only were not included,

42

Excluding the damage-removal treatment


and the poorly designed treatment that did not
ow back, the six successful restimulations and
one uneconomic treatment added incremental
reserves at about $0.20/Mcf. This cost is closer to
the $0.10 to 0.20/Mcf range of past restimulations, even though post-treatment evaluations
indicated that a few pay zones in some of the
wells were not stimulated effectively. Even when
the three unsuccessful treatments are included,
this field trial was highly successful, yielding
additional reserves of 300 MMcf/well [8.6 million
m3/well] at an average cost of $81,600 per well.
There are about 200,000 unconventional gas
wells in low-permeability sands, coalbed
methane deposits and gas shales in the 48 contiguous states of the USA. At least 20%, or about
40,000 wells, could be potential restimulation
candidates. Extrapolating GTI results using the
average incremental recovery of 300 MMcf/well
yields 12 Tcf [343.6 billion m 3] of additional
reserves from refracturing. Companies operating
in the Green River and East Texas formations
continued to perform restimulation treatments
using knowledge gained from this study.

only actual expenses for treatment


implementation. The project team analyzed all
nine wells to better understand each candidateselection method.6
The team considered treatments generating
incremental reserves at a cost of less than
$0.50/Mcf as economic successes. On this basis,
six of the nine wells restimulated at the three
sites were successful (above). All nine wells
combined added 2.9 Bcf [83 million m3] of incremental reserves at a total cost of $734,000, or an
average reserve cost of $0.26/Mcf.

Candidate-Selection Methods
Overall, the GTI refracturing tests were successful, but did not definitively identify a single
candidate-selection method as most effective.
Each technique tends to select different wells
for different reasons that may all be valid,
depending on specific reservoir characteristics
(next page, top). Production statistics worked
reasonably well in the Piceance basin. Virtual
intelligence and pattern recognition worked
best in the Green River basin. Type curves were
most effective in the East Texas basin. Clearly,
additional evaluations were needed to validate
the effectiveness of each technique and to
advance refracturing acceptance.
A reservoir simulation of a hypothetical
tight-gas field was designed for this purpose.7
The objective of this study was to independently
test and validate candidate-selection methods
against the simulation model. Results from this
simulation confirmed that each candidateselection method being studied tended to yield
different candidates. And like the 1998 GTI
restimulation study, some wells were selected by
more than one of the methods. The virtual-intelligence method was generally most effective,
followed closely by type curves. With less efficiency than random selections, production
statistics alone were the least effective method.

Oileld Review

Site,
field/basin

Top 50 candidate-well ranking


Production
Virtual
statistics
intelligence

Success/
failure

Well

Type
curves

Big Piney
and LaBarge/
Green River

GRB 45-12
GRB 27-14
NLB 57-33
WSC 20-09

S
F
F
S

>50
>50
4
38

*15
*39
*>50
*2

>50
32
20
1

Rulison/
Piceance

Langstaff 1
RMV 55-20

S
F

1
43

>50
>50

>50
17

Carthage/
East Texas

CGU 15-8
CGU 3-8
CGU 10-7

S
S
S

>50
>50
4

>50
>50
26

11
7
40

*Revised analysis
Note: Bold italic numbers indicate correct classifications (true positive or true negative)

> Candidate-selection performance. Based on the economic criterion of adding incremental reserves
at less than $0.5/Mcf, the GTI study evaluated the capability of each candidate-selection method to
correctly select successful refracturing candidates or to not select unsuccessful candidates. This
determination was based on whether each method ranked a well among the top 50 candidates or
not. The three methodsproduction statistics, virtual intelligence and pattern recognition, and type
curvesidentied successful refracturing candidates or noncandidates in at least four of the nine
test wells, ve in the case of virtual intelligence. The three methods combined identied only two of
the ve successful treatments and none of the three unsuccessful wells.

Production statistics

Virtual intelligence

14
15
7

50
10

5
103
89

45

53

49
4
93
71

52

120

83

Type curves

6. Ely JW, Tiner R, Rothenberg M, Krupa A, McDougal F,


Conway M and Reeves S: Restimulation Program
Finds Success in Enhancing Recoverable Reserves,
paper SPE 63241, presented at the SPE Annual Technical
Conference and Exhibition, Dallas, Texas, USA,
October 14, 2000.
7. Reeves SR, Bastian PA, Spivey JP, Flumerfelt RW,
Mohaghegh S and Koperna GJ: Benchmarking of
Restimulation Candidate Selection Techniques in
Layered, Tight Gas Sand Formations Using Reservoir
Simulation, paper SPE 63096, presented at the SPE
Annual Technical Conference and Exhibition, Dallas,
Texas, USA, October 14, 2000.

Autumn 2003

< Candidate selection from the GTI


reservoir-simulation study. The top
18 refracturing candidates represent 15% of the wells from the
reservoir stimulation. Virtual intelligence independently selected 10
of the 13 true candidate wells, the
most of any method. These 10
wells consisted of ve that were
uniquely selected by virtual intelligence, one well that was also
selected by production statistics,
two wells that were also selected
by type curves, and two wells that
were selected by all three techniques. The type-curve method
added three true candidate wells
to the combined selections, making the combined number of
correct selections between the virtual intelligence and type-curve
methods 13 out of 13. In practice,
however, no one knows in advance
which wells are true candidates.

8. Emrich C, Shaw D, Reasoner S and Ponto D: Codell


Restimulations Evolve to 200% Rate of Return, paper
SPE 67211, presented at the SPE Production and
Operations Symposium, Oklahoma City, Oklahoma, USA,
March 2427, 2001.
Shaefer MT and Lytle DM: Fracturing Fluid Evolution
Plays a Major Role in Codell Refracturing Success,
paper SPE 71044, presented at the SPE Rocky Mountain
Petroleum Technology Conference, Keystone, Colorado,
USA, May 2123, 2001.
Sencenbaugh RN, Lytle DM, Birmingham TJ,
Simmons JC and Shaefer MT: Restimulating Tight Gas
Sand: Case Study of the Codell Formation, paper SPE
71045, presented at the SPE Rocky Mountain Petroleum
Technology Conference, Keystone, Colorado, USA,
May 2123, 2001.

The first stage of the 1998 GTI study and


results from this simulation provided valuable
insights into the effectiveness of each candidateselection methodology, but each technique
needed to be tested using real eld data. Rather
than establish a new database of restimulation
cases for this purpose, as was the original overall
project objective, participants in the 1998 GTI
study sought a eld with a history of restimulation activity and results. With an existing
dataset, the approach used for the simulator
study could be repeated in an actual eld setting
to evaluate each candidate-selection method.
As follow-up to the reservoir simulation, GTI
selected the Wattenburg eld to further evaluate
candidate selection methods using actual field
data. This tight-gas development, located north
of Denver, Colorado, on the western edge of the
Denver-Julesburg basin, was attractive because
more than 1500 area wells had been refractured
since 1977. Most of these treatments were economically successful.8
Patina Oil & Gas Corporation, a leading
operator in this basin, had performed about
400 fracture restimulations from 1997 through
2000, and agreed to participate. This allowed a
candidate-selection algorithm developed
independently by Patina to be used in addition to
the three GTI candidate-selection methods.
The methods were evaluated without disclosing beforehand those wells that had actually
responded favorably to restimulation. Afterward,
candidate selections were compared with actual
well performance. This approach allowed the
effectiveness of each method to be assessed.
Candidate selection using actual Wattenburg
field data confirmed previous GTI study and
reservoir-simulation results.
Prioritizing refracturing candidates provides
considerable value during restimulation
programs. In the absence of prior restimulation
results, both pattern recognition and type curves
are useful for selecting restimulation candidates; production statistics are less effective.
Virtual intelligence and other patternrecognition techniques, which use prior
refracturing data and results to learn from,
can further improve candidate selection and
restimulation success. The GTI field trials,
reservoir simulation and Wattenburg field
evaluation confirmed that the performance of
each candidate-selection method appeared to be
reservoir specic (bottom left).

43

Analysis of production statistics tends to


identify completions that underperform
compared with offset wells. Substandard performance could result from a poor quality reservoir,
but this method should be valid in fields with
relatively uniform reservoir quality and fairly
stable production.
Virtual-intelligence methods tend to select
wells that have less than optimal original
completions or stimulation procedures. Patternrecognition technologies should be applied
when reservoir, completion and stimulation
complexity is high.
Type curves tend to identify candidate wells
based solely on incremental production potential, and therefore, weights the better producing
wells in a eld more heavily. This method should
be used when production data quality is good and
petrophysical information is readily available.
The applicability of any candidate-selection
process should be assessed for each specific
area being evaluated. In effect, an ideal
methodology may combine several techniques.
The three efforts to evaluate candidate-selection
methods also indicated that nonanalytical
analyses, such as evaluating current producing
rate and estimated ultimate recovery to identify
underperforming wells, could be useful for
candidate selection in the absence of
other approaches.
A Field-Wide Evaluation
Prior to 1999, refracturing by Patina Oil & Gas
Corporation in the Wattenburg eld had primarily targeted underperforming wells and
completions that screened out prematurely or

had mechanical failures during the initial stimulation. When other operators began restimulating
their better producers with varying, but generally
encouraging results, Patina initiated a eld-wide
evaluation of refracturing potential.
The Wattenburg field produces mainly from
the Codell interval. This ne-grained sandstone,
deposited in a marine-shelf environment, is a
member of the Upper Cretaceous Carlisle shale.
The Codell reservoir contains 15 to 25% clay by
volume in mixed layers of illite and smectite
that ll and line the pore spaces.
The pay interval is 14 to 35 ft [4.3 to 10.7 m]
thick, 6800 to 7700 ft [2073 to 2347 m] deep and
continuous across the field. Permeability is
less than 0.1 mD. Porosity from density logs is
8 to 20%. Initially, the reservoir was overpressured with a gradient of about 0.6 psi/ft
[13.5 kPa/m]. Bottomhole temperature is 230 to
250F [110 to 121C]. Wells are drilled on a
40-acre [162,000-m2] spacing.
During 1998, Patina compiled a database of
250 fracture restimulations on both operated
and nonoperated properties. After eliminating
wells treated with borate crosslinked fluids,
which were 20% less productive than other
wells, company engineers focused on the
remaining 200 wells. These wells had been restimulated with carboxymethyl hydropropyl guar
(CMHPG) or hydropropyl guar (HPG) uids.
Further evaluation identified 35 discrete
geologic, completion and production parameters
related to well performance. Linear-regression
analysis helped determine those parameters
that correlated with peak incremental production after refracturing. Two technical

improvements from this field-wide evaluation


provided an order-of-magnitude improvement in
restimulation results.
The first was application of carboxymethylate guar (CMG) fluids with lower polymer
loadings, which maintain proppant transport
and minimize residual proppant-pack damage
from unbroken and unrecovered gel. Nondamaging fluids are particularly important in the
refracturing of low-permeability formations
where long-term gas saturation has been established and reservoir pressure may be depleted.
The second improvement was a candidateselection method developed by Patina that uses
historical restimulation results in the basin.
Together with CMG fluids, this statistically
based algorithm achieved significant improvements in selection of the best refracturing
candidates (below). Average peak incremental
production rate almost doubled from just over
1000 to about 2000 barrels of oil equivalent
(BOE)/well/month [159 to 318 m3/well/month],
which equaled about 80% of the average initial
production rate. The associated rate-of-return
on refracturing investments increased from 66%
to more than 200% at $2.50/Mcf. Estimated
incremental recoveries increased from 25 to 38
million BOE per well [4 to 6 million m3/well].
Only about 3% of refracturing treatments
resulted in economic failures, primarily because
the propped fractures communicated with the
overlying Niobrara formation or an offset well.
This failure rate may become higher as refracturing density increases. The overall success of
this program resulted from stringent
well-selection criteria, strict quality-control

2500

Peak production, BOE/well/month

Development and application of genetic


algorithm for candidate selection
2000

1500

1000
Patina
500
Others
CMG fluids
0
1997

1998

1999

2000

> Historical refracturing performance in the Wattenburg eld, Colorado. The combined applications of CMG stimulation uids and
the candidate-selection algorithm developed by Patina Oil & Gas signicantly improved restimulation results in Patina-operated wells.

44

Oileld Review

Description

Statistical
significance

Rank

Parameter

Hydrocarbon volume,
porosity-feet

Net pay for Codell above a


10% density porosity cutoff

38%

Cumulative
recovery factor

Cumulative gas recovered divided


by original gas in place (OGIP) for
40-acre drainage area

17%

Initial completion

Peak rate premium assigned


if well was originally completed
limited entry in Codell-Niobrara

9%

Estimated ultimate
recovery (EUR) factor

EUR divided by OGIP for 40-acre


drainage area

11%

Gas/oil ratio

Projected ultimate gas/oil ratio

20%

Maximum differential
recovery, million BOE

EUR difference between subject


well and best offset well within
one mile of subject well

5%

> Patina Oil & Gas statistical algorithm. Of the ve statistically signicant
variables of the candidate-selection algorithm for Wattenburg eld, hydrocarbon volume in porosity-feet represents reservoir quality, initial
completion represents the initial completion, and the other threecumulative recovery factor, estimated ultimate recovery factor and gas/oil
ratiorepresent well performance. Well location is not signicant because
of the relatively uniform reservoir quality. However, higher, and therefore better, gas/oil ratios do tend to occur in the center of the eld. The sixth variable
maximum differential recovery in BOE helps predict restimulation potential
for economic evaluations.

guidelines for treatment fluids and effective


operational practices in the eld.
Other area operators have reported similar
improvements in productivity, economic results
and recovery from refracturing.9 Based on these
results, more than 4000 other wells in the
Piceance basin may be candidates for restimulation. Patina engineers continue to expand their
already extensive refracturing database and netune the candidate-selection algorithm. In some
wells, Patina and other area operators are now
successfully fracturing wells for a third time.
Candidate-Selection Criteria
The Patina Oil & Gas linear-regression analysis
identified five statistically significant variables
that were incorporated into the Wattenburg eld
candidate-selection algorithm (above). Although
statistically less significant, a sixth variable
maximum differential recovery in BOE, was
added to help predict restimulation results for
economic evaluation purposes.
Hydrocarbon pore volume, or porosity-feet,
the most statistically significant parameter, is
incorporated in the cumulative and ultimate

Autumn 2003

recovery factors. Gas/oil ratio, which varies


from about 5000 to 35,000 scf/bbl [900 to
6304 m3/m3], correlates to higher recovery wells
from original and refractured completions
primarily in and around central areas of the
eld. This is indicative of greater relative permeability to gas because formation thickness and
reservoir permeability are relatively uniform
across the eld.
Well completions that used limited-entry
perforating across both the Codell and Niobrara
formations resulted in shorter effective fracture
lengths in the Codell than those completed only
in the Codell. Cumulative and ultimate recovery
factors determined from individual well and
reservoir parameters coupled with decline-curve
analysis indirectly represented the extent of
depletion and the capability of the reservoir to
ow back and clean up treatment uids. These
factors also provided an indication of whether
new hydraulic fractures might reorient with
respect to the original propped fracture (see
Fracture Reorientation, page 47).
The maximum differential BOE is the difference in ultimate recovery between the subject

well and the best well within 1 mile [1.6 km].


This parameter gives an indication of upside
reserve potential in the immediate vicinity of a
subject well. Engineers eliminated some variables, such as faulting, treatment size and
perforated interval, which were statistically
insignificant. Well location is not significant in
this field because of the relatively uniform
reservoir quality.
Post-refracturing performance continues to
support added reserves above baseline projections for the original completions because the
initial completion in most of the wells was not
effectively draining the 40 acres allotted to each
well in the development pattern. A reevaluation
of 1000 refracturing treatments indicated good
correlation with the best t of actual results. To
some extent, these variables can be quantified
for individual wells by analyzing actual production in terms of long-term pressure drawdown
using production type-curve analysis techniques.
Production type-curve analysis requires more
analysis time, but effectively forecasts restimulation results with a higher degree of accuracy
than do other statistical techniques.
Variations still existed, but overall the Patina
algorithm successfully ranked restimulation
potential on a eld-wide basis. The variability in
refractured well performance appears to result
from an inability of statistical methods to
differentiate between actual drainage areas,
differences in matrix permeability, effective
fracture lengths from the original stimulation
and the impact of liquid condensate loading, or
banking, around these wellbores using only
production and completion parameters.10
The fundamental objective of refracturing is
to enhance well productivity. However, restimulation is viable only if wells are underperforming
because of completion-related problems, not
because of poor reservoir quality. Neither fracturing nor refracturing can turn marginal
producers in poor reservoirs into good wells. To
prioritize and select refracturing candidates,
engineers must understand the reasons for poor
performance in previously fractured wells.
9. Shaefer and Lytle, reference 8.
Sencenbaugh et al, reference 8.
10. Barnum RS, Brinkman FP, Richardson TW and
Spillette AG: Gas Condensate Reservoir Behaviour:
Productivity and Recovery Reduction Due to
Condensation, paper SPE 30767, presented at the SPE
Annual Technical Conference and Exhibition, Dallas,
Texas, USA, October 2225, 1995.

45

Ineffective or problematic
initial completions
. Unstimulated horizons
. Low fracture conductivity
. Short fracture length
. High skin, or damage

Gradual formation damage


during production
. Scale and fines
. Workover frequency
. Well age

Well
underperformance

Technology evolution
. Advanced stimulation technology
. New completion techniques
. Well age

> Potential causes of underperformance in previously stimulated wells. The


GTI restimulation project team established a classication framework to help
diagnose problems in hydraulically fractured wells that perform below operator
expectations. At the highest level, there are three broad categories: ineffective
or problematic initial completions, gradual production damage and advances
in technology or evolving techniques compared with past practices.

Completion-Related Underperformance
To aid in problem diagnosis, the 1998 GTI
project established a framework to classify
well-performance problems (above). For tightgas wells, three specific problems, ranked in
order of highest perceived restimulation potential were identied:
Unstimulated or bypassed pay
Insufcient fracture conductivity
Insufcient fracture length.
Ineffective or problematic initial completions are the most common type of problem.
Examples include lack of quality control during
initial fracture treatments, residual polymer
damage from stimulation fluids, inappropriate
proppant selection, premature screenout, underdesigned fracturing treatments, incompatible
fluids and single-stage treatments that leave
some pay intervals unstimulated.
Hydraulic fractures can lose effectiveness in
the years after an initial stimulation treatment
because of gradual damage that occurs over the
life of a well. Examples include loss of fracture
conductivity from proppant crushing or embedding in the formation and plugging of the pack
by formation nes or scale deposition. Proppant
owback from the near-well area can allow the
hydraulic fractures to close. Typically, little
information is available to identify these
specic mechanisms.
Wells with these types of problems have the
greatest potential for remediation by refracturing. In older wells that have a higher occurrence
of these problems, reservoir pressure must be
sufcient to justify refracturing, both in terms of

46

remaining reserves and adequate flowback of


treatment uids. Well age may be the best indicator of gradual damage and the possibility of
applying new stimulation technology.
Diagnosing production damage, a second
major category of problems, often is difficult.
Proppant owback, uid damage and high skin
factors, frequent remedial workovers, and nes
or scale deposits during the onset of multiphase
flow or water breakthrough are manifestations
of problems that develop over time. Any
combination of these may indicate that well productivity has deteriorated over time.
A third category, advances in completion and
stimulation technology, also provides opportunities to restimulate wells originally completed
using older technology. New treatment designs,
advanced computer models, less damaging
fracturing fluids, improved fluid additives and
proppants help create longer, wider, more
conductive fractures. In some sense, this
category is a subset of the previous two because
older technology often is synonymous with less
effective initial completions where more gradual
damage has occurred.
It is important to determine what types of
productivity problems correlate with the best
refracturing candidates in a eld, area or basin.
Engineers can gain information about specific
well-completion problems and how to remediate
them by reviewing individual well records.
Unstimulated zones typically result from
using limited-entry diversion or from fracturing
multiple pay horizons in a single-stage treatment. This well-completion problem may

represent the greatest restimulation potential for


two reasons. First, tight-gas wells are frequently
multiple-zone completions. The tendency is to
treat multiple intervals in fewer stages to reduce
treatment cost. Second, enhanced well productivity from stimulation of new zones almost
always represents an incremental reserve addition, not just an increase in production rate and
accelerated reserve recovery.
A low ratio of fracture-treatment stages
and proppant volume to the number and distribution of net-pay intervals is an indicator of
potentially understimulated or unstimulated
zones. Radioactive tracer surveys, well tests,
production-decline curves and production logs
also help diagnose unstimulated or poorly
performing intervals.
Insufficient conductivity of an initial
propped fracture probably represents the next
highest restimulation potential. However, the
distinction between rate acceleration and true
incremental reserve addition from increased
conductivity after refracturing is often blurred.
Examples include insufcient proppant strength
for the fracture-closure pressure at reservoir
depth, proppant settling, low proppant concentrations and damage to proppant packs by
partially broken and unbroken gel.
Capturing incremental reserves at the outer
margin of a drainage area by increasing fracture
length is difficult. A relatively small treatment
compared with the higher net-pay thickness is
generally indicative of limited fracture length.
Generating longer hydraulic fractures can be
expensive unless the initial treatment was
extremely small. However, if restimulation
achieves additional fracture length and expands
the drainage area of a well, incremental production should represent a true reserve addition.
A review of the initial fracturing treatment
and flowback helps identify possible limited
fracture conductivity and length. Well-test
and production-decline analyses also help diagnose these conditions. A short period of linear
flow followed by radial flow after fracturing
indicates insufficient fracture conductivity or
inadequate length.
Refracturing opportunities also exist as a
result of eld development and well production
provided wells have enough pressure to flow
back and produce, even if energized treatment
uids or articial lift is required. In addition to
lower pore pressure, pressure depletion also
implies higher effective stress, which results in
less hydraulic fracture width and longer lateral
extension for the same volumes of treatment
uid and proppant.

Oileld Review

In addition, depletion of pay intervals


increases the stress contrast between pay intervals and bounding shales, which improves
vertical containment and allows generation
of longer fractures. Alteration of horizontal
in-situ stress around a wellbore and an existing
fracture also may contribute to fracture reorientation during restimulation.
Fracture Reorientation
Historically, refracturing has been a remedial
measure performed on poorly producing wells
with short or low-conductivity initial fractures.
However, there are numerous examples of
successful restimulations on previously fractured wells, especially tight-gas wells, that still
exhibit linear flowa negative 0.5 slope on
log-log production-rate plots indicative of deeply
penetrating, highly conductive fractures. Production tests and history matching using a
numerical simulator that accommodated orthogonal fractures and horizontal permeability
anisotropy indicate a strong probability of
refracture reorientation in many of these wells.
This concept of fracture reorientation is not
new and has been modeled in full-scale
laboratory experiments. In addition, fracture
reorientation has been observed in soft, shallow
formations.11 After an initial period of production, stress changes around existing wells with
effective initial fracture treatments may allow
new fractures to reorient and contact areas of
higher pore pressure.
Laboratory tests have also shown that matrix
pore-pressure changes inuence hydraulic fracture orientation in the reservoir volume between
injectors and producers in a waterflood.12 The
fractures orient normal, or perpendicular, to the
higher stress gradient. Fractures initiated from
producing wells orient towards and intersect
the injection well if the stress gradient is
high enough and the in-situ stress anisotropy is
not dominant.
Pressure changes around a deeply penetrating, highly conductive fracture also create high
stress gradients normal to the initial fracture
that may cause fracture reorientation during
restimulation treatments. Stress changes reach
a maximum and then diminish with further
depletion. An optimal window of time during
which to perform refracturing treatments can be
determined.13 Horizontal permeability anisotropy
further increases these stress changes. Similarly,
a separate study showed that initial fracture
orientation is influenced by production in
unfractured formations that have large horizontal permeability anisotropy.14

Autumn 2003

GTI provided funding for Schlumberger to


investigate these concepts in greater detail.15
Numerical simulations during this investigation
provided evidence that new fractures can form
at angles up to 90 from the initial propped
fracture azimuth (below). Fracture reorientation bypasses damage caused by drilling and
completion activities, and avoids zones of
reduced permeability caused by compaction and
other flow restrictions, including hydrocarbon
liquid dropout, or condensate banking, around
a well.
The horizontal stress component parallel to
an initial fracture is reduced more quickly as a
function of time than the perpendicular component. If these induced stress changes overcome
the original stress differential, then a new fracture will initiate and propagate along a different

azimuthal plane than the initial fracture until it


reaches the boundary of the elliptical stressreversal region. The fracture may continue along
the new azimuth for some distance beyond this
point, depending on formation toughness.
Many factors contribute to the location of
the stress-reversal boundary, including production history, reservoir permeability, fracture
dimensions, pay-zone height, elastic properties
of the pay and bounding barrier zones, and the
initial horizontal stress contrast. These parameters can be modeled and should be considered
when selecting refracturing candidates.
Computer simulations can determine the
optimal time window for refracturing and
fracture reorientation. Wells with long initial
fractures in low-permeability formations have a
longer time window. Production shut-in periods

y
New fracture
Isotropic point
Stress-reversal
region
Maximum
horizontal
stress

Wellbore
x

Initial fracture
Isotropic point

New fracture

Minimum
horizontal
stress

> Stress reorientation and orthogonal fracture extension. This horizontal section through a vertical wellbore depicts an original hydraulic fracture in the
x direction and a second reoriented fracture in the y direction. Fluid production after placement of the initial fracture can cause a local redistribution
of pore pressure in an expanding elliptical region around the wellbore and
initial fracture. The stress-reversal boundary is dened by isotropic points of
equal primary horizontal stresses. Stress reorientation and fracture extension
in a direction away from the initial propped fracture help explain pressure
responses during refracturing treatments and unanticipated production
increases from refractured wells known to have effective initial fractures.

11. Wright CA, Stewart DW, Emanuel MA and Wright WW:


Reorientation of Propped Refracture Treatments in the
Lost Hills Field, paper SPE 27896, presented at the SPE
Western Regional Meeting, Long Beach, California, USA,
March 2325, 1994.
Wright CA, Conant RA, Stewart DW and Byerly PM:
Reorientation of Propped Refracture Treatments,
paper SPE 28078, presented at the SPE/ISRM Rock
Mechanics in Petroleum Engineering Conference,
Delft, The Netherlands, August 2931, 1994.
Wright CA and Conant RA: Hydraulic Fracture
Reorientation in Primary and Secondary Recovery from
Low-Permeability Reservoirs, paper SPE 30484, presented at the SPE Annual Technical Conference and
Exhibition, Dallas, Texas, USA, October 2225, 1995.
12. Bruno MS and Nakagawa FM: Pore Pressure
Inuence on Tensile Propagation in Sedimentary Rock,
International Journal of Rock Mechanics and Mining
Sciences and Geomechanics Abstracts 28, no. 4
(July 1991): 261273.

13. Elbel JL and Mack MG: Refracturing: Observations


and Theories, paper SPE 25464, presented at the SPE
Production Operations Symposium, Oklahoma City,
Oklahoma, USA, March 2123, 1993.
14. Hidayati DT, Chen H-Y and Teufel LW: Flow-Induced
Stress Reorientation in a Multiple-Well Reservoir,
paper SPE 71091, presented at the SPE Rocky Mountain
Petroleum Technology Conference, Keystone, Colorado,
USA, May 2123, 2001.
15. Siebrits E, Elbel JL, Detournay F, Detournay-Piette C,
Christianson M, Robinson BM and Diyashev IR:
Parameters Affecting Azimuth and Length of a
Secondary Fracture During a Refracture Treatment,
paper SPE 48928, presented at the SPE Annual Technical
Conference and Exhibition, New Orleans, Louisiana,
USA, September 2730, 1998.

47

should be minimized to maintain a high porepressure gradient normal to the initial fracture.
Aside from this, standard fracture design considerations should be used.
Fracture restimulations in the naturally fractured Barnett Shale, north of Fort Worth, Texas,
USA, are an example of fracture reorientation.
These treatments were monitored with an
array of surface and subsurface tiltmeters
(below).16 The results suggested signicant fracture reorientation in one well and oblique
reorientation in the other well. Post-treatment
production increased substantially in both wells.
Other refractured wells in the area had similar
increases. Reservoir depletion combined with
natural fractures can cause complex fracture
networks to develop during initial treatments
and restimulations.

A Gas-Shale Restimulation Program


In 1997, Mitchell Energy, now Devon Energy,
began using greatly reduced polymer concentrations in treatment fluidscurrently only
surfactant-base friction-reducing agents are
usedand much lower volumes of proppant in
the Barnett Shale formation. These slick-water
fracturing treatments have been extremely
successful and are similar to designs used by
operators for Cotton Valley sandstone stimulation treatments in the nearby East Texas basin.
Additional gas-shale development efforts are
currently under way in other areas of North and
West Texas. The Barnett Shale, for example, is
present in wells from the Fort Worth basin to
the Permian Basin of West Texas, so lessons
learned in North Texas can be applied in thousands of wells.

Initial fracture azimuth


W

E
Initial injection
1st 83 minutes
2nd 83 minutes
3rd 83 minutes
Final 83 minutes
S

Fracture-induced
surface trough

Depth

Surface tiltmeters

Fracture

Downhole
tiltmeters in
offset well

> Formation displacement around a vertical hydraulic fracture. Extremely


sensitive tiltmeters placed in a radial pattern on the surface around a stimulation well candidate (bottom) can monitor fracture azimuth during stimulation
treatments (top). Fracture geometry is inferred by measuring induced rock
deformations. The deformation eld, which radiates in all directions, can also
be measured downhole by wireline-conveyed tiltmeter arrays in offset wells.

48

Deposited in a deep marine environment, the


Barnett Shale consists of layered mudstone, siltstone and some interbedded limestone with
open and calcite-lled natural fractures. Matrix
permeability in this rich organic, fine-grained,
Mississippian-age shale formation is extremely
low, about 0.0001 to 0.001 mD. Estimated
ultimate recovery for a typical Barnett Shale
well is 0.5 to 1 Bcf [14.3 to 28.6 million m3]. This
represents a calculated recovery of 8 to 10 % of
the gas in place. Achieving economic production
requires large fracturing treatments.
The Barnett Shale typically lies between the
upper Marble Falls limestone and the lower Viola
limestone. In some areas, the Viola formation is
replaced by the Ellenburger dolomite, which is
not as competent as the Viola for confining
hydraulic fractures. The Barnett Shale is 200 to
1000 ft [61 to 305 m] thick, averaging about
500 ft [152 m] in the main area of the eld.
In 1999, analysis of near- and far-stress elds
in the Barnett determined that new fractures
created during restimulation followed the
original fracture plane for a short distance
before taking a new direction.17 Recent microseismic surveys conducted during refracturing
treatments confirm that new fractures propagate initially in the original northeast-southwest
direction before diverging along a new northwest-southeast azimuth (next page, top).18 In
addition to fracture reorientation, microseismic
mapping, such as StimMAP hydraulic fracture
stimulation diagnostics, also provide evidence of
complex fractures that contribute further to
increased well productivity from the Barnett
Shale (next page, bottom).
Infill wells drilled on a spacing as close as
27 acres [109,300 m2] indicated long elliptical
drainage patterns. Refracturing, therefore,
offers significant potential for increased well
productivity and improved gas recovery by creating new fractures that contact other areas of the
reservoir as a result of fracture reorientation
and creation of complex hydraulic fracture
networks. Restimulations also address
underperformance caused by ineffective well
completionsprimarily early termination of the
initial treatmentbypassed or unstimulated
zones and gradual production damage in this
naturally fractured formation.
Barnett Shale completions date back to the
1980s, when acid breakdown and fracturing
treatments used high polymer concentrations,
crosslinked-gel fluids and moderate proppant
concentrations with minimal external gel
breaker because of high formation temperatureabout 200F [93C]. Some of the initial

Oileld Review

Microseism

Receivers

Reservoir
Fracture
Wellbore

Offset
wellbore

> Microseismic fracture mapping. Microseismic imaging relies on detection


of microearthquakes or acoustic emissions associated with hydraulic fracturing or induced movement of preexisting fractures. This technique uses
three-component sensors, typically 5 to 12 geophones or accelerometers, in
an offset observation well to detect these extremely small events, or microseisms. Normally, perforating operations in the well being monitored are used
to calibrate and orient the sensors. As a treatment proceeds, the microseisms
generated by fracture propagation are detected, oriented and located with
the reservoir to develop a fracture map.

Simple fracture

Complex fractures

Extremely complex fractures

> Complex fracture networks. The simple classical description of a hydraulic fracture is a single,
biwing, planar crack with the wellbore at the
center of the two wings (top). In some formations,
however, complex (middle) and very complex
(bottom) hydraulic fractures may also develop, as
appears to be the case in the naturally fractured
Barnett Shale.

16. Siebrits E, Elbel JL, Hoover RS, Diyashev IR, Grifn LG,
Demetrius SL, Wright CA, Davidson BM, Steinsberger NP
and Hill DG: Refracture Reorientation Enhances
Gas Production in Barnett Shale Tight Gas Wells, paper
SPE 63030, presented at the SPE Annual Technical
Conference and Exhibition, Dallas, Texas, USA,
October 14, 2000.
Fisher MK, Wright CA, Davidson BM, Goodwin AK,
Fielder EO, Buckler WS and Steinsberger NP: Integrated Fracture Mapping Technologies to Optimize
Stimulations in the Barnett Shale, paper SPE 77441,
presented at the SPE Annual Technical Conference
and Exhibition, San Antonio, Texas, USA,
September 29October 2, 2002.

Autumn 2003

Maxwell SC, Urbancic TI, Steinsberger N and Zinno R:


Microseismic Imaging of Hydraulic Fracture Complexity
in the Barnett Shale, paper SPE 77440, presented at the
SPE Annual Technical Conference and Exhibition, San
Antonio, Texas, USA, September 29October 2, 2002.
17. Siebrits et al, reference 16.
18. Fisher et al, reference 16.
Maxwell et al, reference 16.
19. Willberg DM, Steinsberger N, Hoover R, Card RJ and
Queen J: Optimization of Fracture Cleanup Using
Flowback Analysis, paper SPE 39920, presented at
the SPE Rocky Mountain Regional/Low-Permeability
Reservoirs Symposium and Exhibition, Denver, Colorado,
USA, April 58, 1998.

treatments also included CO 2 or N 2 . Initial


post-treatment production increases were
encouraging, but short-lived. These practices
continued through 1990.
Early treatments yielded poor fracture
conductivity because of damage caused by
incomplete treatment-uid cleanup and polymer
degradation, and by the ne silica our used as
a fluid-loss additive, which remained in the
proppant pack. Shorter fracture length resulted
from small treatment volumes. Data from
production logs indicated that some sections of
the Barnett remained untreated or understimulated and provided little or no gas production
after initial fracturing treatments.
Gradual completion damage and productivity
degradation potentially result from insufficient
initial fracture length, incomplete treatmentfluid cleanup and relative-permeability
restrictions caused by water influx from lower
formations. In some wells, there is evidence of
scale deposition when water from incompatible
sources is used in stimulation treatments. Productivity degradation also occurs as reservoir
energy decreases. NODAL production system
analysis indicates that below about 400 Mcf/D
[11,455 m3/d], high fluid levels in the wellbore
restrict gas production. Artificial-lift methods
help increase gas output.
After 1990, operators began reducing polymer concentrations, using N 2 for flowback
assistance, increasing overall uid and proppant
volumes, and pumping maximum sand concentrations of three pounds of proppant added
(ppa) per 1000 gal [360 kg of proppant added
(kgpa) per m3]. These changes were in response
to earlier limited well productivity and disappointing stimulation results. Engineers increased
the use of external breaker systems, eventually
eliminating N 2 and solid fluid-loss additives,
such as ne silica our. Incremental production
from fracture stimulations continued to improve
as a result of these trends in treatment optimization, which culminated in the advent of
slick-water treatments in 1997.
Operators also began to focus on improving
post-treatment cleanup. Previous procedures
were conservative, with limited flowback rates
and treatment cleanup periods that lasted 7 to
10 days. The new procedures reflected a more
aggressive attempt to force fracture closure and
recover as much treatment uid as possible in 2
to 3 days.19
The evolution of fracturing practices from
crosslinked gels to slick water and improved
procedures for treatment-fluid recovery significantly enhanced gas production from the

49

100,000

Typical Barnett Shale


restimulation results
Gas rate, Mcf/month

Barnett Shale. Refracturing with large fluid


volumes and lower volumes of proppant yielded
well productivities that, in some cases, are the
highest ever in these wells (right).
It appears that reduction and eventual elimination of solids in fracturing uids generate better
production results in tight-gas formations. Slickwater treatments are currently the accepted
practice for completing new wells and refracturing existing completions in the Barnett Shale. The
reasons for success of this method are not fully
understood and are still under study. One possibility may be that fracture facies do not heal, or
close, completely once displaced or may be etched
and eroded by large stimulation treatments.
Advanced well logs from tools, such as the
FMI Fullbore Formation MicroImager and DSI
Dipole Shear Sonic Imager tools, used in conjunction with standard well-logging suites
provide more detailed formation evaluation and
reservoir characterization. Stress profiles from
sonic logs assist in design and implementation
of multistage treatments to ensure complete
zonal stimulation coverage. The higher level of
detail resulted in additional improvement in
Barnett Shale completions, including more accurate perforation placement across intervals with
identied open natural fractures.

10,000
Refractured

1000

100
1990

1991

1992

1993

1994

1995

1996 1997
Year

50

1999

2000

2001

2002

2003

> Typical restimulation results for a Barnett Shale well. The use of substantial volumes of slick water
and low quantities of proppant sand to refracture the Barnett Shale resulted in well productivities as
good as or better than the original completion. In some cases, the well productivities after refracturing
were the highest ever recorded in this eld.

R14

A Shallow-Gas Restimulation Program


Enerplus Resources Fund realized an average
sixfold increase in production from refracturing
shallow-gas wells in the Medicine Hat and Milk
River formations of southeastern Alberta,
Canada. These results were obtained in a 15-well
stimulation program during the second half of
2002. Ten treatments were performed using the
CoilFRAC stimulation through coiled tubing
service. 20 The CoilFRAC technique utilized a
straddle isolation tool that allowed individual
perforated intervals to be selectively isolated
and stimulated. Jointed pipe and a snubbing
unit were used in place of coiled tubing (CT) on
the other five wells. These CT-conveyed and
snubbing-conveyed stimulations helped optimize
fracture treatments and facilitated completion
and stimulation of bypassed zones.
Initially completed in the 1970s, vertical
wells in the Medicine Hat and Milk River formations produce from depths of 300 to 500 m [984
to 1640 ft]. Producing intervals consist of layered
sandstones with high shale content that fracture
easily. These wells were originally fractured by
pumping uids and proppants down casing in a
single-stage operation with ball sealers to divert
the treatment across multiple sets of perforations. To select restimulation candidates,
engineers sought a relationship between initialfracture effectiveness and current production.

1998

R13W4

50.8

223.9

137.3

T20

397.4

570.0

743.1

310.4 483.5 656.6


Cumulative gas, MMscf

916.2

829.6

T20

T19

T19

T18

T18

R14

R13W4

> Shallow-gas restimulation criteria. Because pressure-transient testing and


analysis were too expensive and not economically practical for this project,
Enerplus Resources Fund chose production data as the best relative indicator
of gradual damage, connectivity and initial stimulation effectiveness. Cumulative gas production data were contoured and color-coded using gas-mapping
software. This allowed engineers to easily identify and select refracturing
candidates in areas with lower recovery factors (blue).

Oileld Review

casing scraper was run on all wells to clear the


wellbore of restrictions and to verify the minimum internal diameter.
Intervals targeted for restimulation were
reperforated to ensure injectivity and improve
treatment effectiveness. Because of a lack of upto-date logs, existing intervals were reperforated
at the same depths and lengths as the initial
perforations. Pretreatment well evaluations confirmed interval lengths and sand quality from
gamma ray logs. In four wells stimulated through
coiled tubing, additional net-pay intervals were
perforated based on existing logs.
Cumulative production and current producing rates proved effective in selecting
restimulation candidates. Refracturing resulted
in an average per-well production increase of
about six times the prestimulation rate. Six of
the 15 wells had higher average post-fracture

Average production rate for CoilFRAC restimulations

rates than at the time of initial completion; four


wells produced within 25% of their original
three-month completion rates in the 1970s.
This substantial level of productivity increase
is even more impressive when viewed in the
context of almost 30 years of production and
more than 100 psi [689 kPa] of pressure depletion (below).
These results are consistent with documented
evaluations of other CoilFRAC treatments
performed in the area since 1997. 21 Average
production from wells fractured through coiled
tubing was slightly higher than treatments
performed with a snubbing unit. This further
confirms that fracturing many small intervals
yields better production rates than fracturing a
few larger intervals. In addition, coiled tubingconveyed fracturing costs about 10% less than
snubbing-unit treatments.

Average production rate for snubbing-unit restimulations


140

450 psi

120

335 psi

100

Pressure depletion
over 30 years

80
60
40

335 psi

20

Average production rate, Mcf/D

140

Average production rate, Mcf/D

These wells were completed initially within a


two-year period, so cumulative production is
normalized over 30 years. Analysis indicated
that average production in the first three
months after initial completion was directly
proportional to the 30-year cumulative gas
production. Furthermore, gas rates and stimulation effectiveness are related, so stimulation
effectiveness is directly proportional to cumulative production.
Completions with lower cumulative gas production than nearby wells were identified as
candidates for refracturing (previous page, bottom). Other considerations included average
production in the rst three months after initial
completion, productive interval lengths, vertical
distance between perforated intervals and current production rate. Wells producing at
currently economical rates of more than 25
Mcf/D [716 m3/d] were eliminated as refracturing candidates.
Intervals greater than 7 m [23 ft] were eliminated as CoilFRAC candidates. Snubbing-unit
operations allowed longer straddle-tool isolation
lengths up to about 15 m [49 ft]. Additionally,
because of the risk of fractures growing vertically
into adjacent intervals, intervals closer together
than about 10 m [33 ft] also were eliminated.
The length of individually perforated zones
fractured with coiled tubing varied from 0.9 m to
6.1 m [3 to 20 ft] with four to seven zones
treated in each well. Zones fractured using the
snubbing technique varied from 3 m to 14 m [9.8
to 45.9 ft] in perforated length. The number of
zones treated ranged from two to four zones
per well.
Because of the age of these wellbores,
precautions were taken to avoid potential
mechanical failures. Surface casing vent flows
were checked; any indication of gas migration to
surface eliminated the well as a candidate. A

120

450 psi
335 psi

100

Pressure depletion
over 30 years

80
60
40

335 psi

20
0

Well life
Initial

Well life
Before refracturing

After refracturing

Field production
5.0
6 new wells

4.5

Coiled tubing cleanout


of new wells

Autumn 2003

Production, MMscf/D

4.0
20. Degenhardt KF, Stevenson J, Gale B, Gonzalez D, Hall S,
Marsh J and Zemlak W: Isolate and Stimulate
Individual Pay Zones, Oileld Review 13, no. 3
(Autumn 2001): 6077.
21. Lemp S, Zemlak W and McCollum R: An Economical
Shallow-Gas Fracturing Technique Utilizing a Coiled
Tubing Conduit, paper SPE 46031, presented at the
SPE/ICoTA Coiled Tubing Roundtable, Houston, Texas,
USA, April 1516, 1998.
Zemlak W, Lemp S and McCollum R: Selective
Hydraulic Fracturing of Multiple Perforated Intervals
with a Coiled Tubing Conduit: A Case History of the
Unique Process, Economic Impact and Related
Production Improvements, paper SPE 54474, presented
at the SPE/ICoTA Coiled Tubing Roundtable, Houston,
Texas, USA, May 2526. 1999.
Marsh J, Zemlak WM and Pipchuk P: Economic
Fracturing of Bypassed Pay: A Direct Comparison of
Conventional and Coiled Tubing Placement Techniques,
paper SPE 60313, presented at the SPE Rocky Mountain
Regional/Low Permeability Reservoirs Symposium,
Denver, Colorado, USA, March 1215, 2000.

Two out of five


snubbing-unit
refractured wells
on-line

3.5
3.0

13 new
wells

2.5
2.0
Last well to be CT
fractured (only 10 of
15 wells have been
fractured at this point
and all through CT)

1.5
1.0
Gas compressor
shutdown

0.5
0.0
2001

2002

> Shallow-gas restimulation results. Refracturing shallow wells in the gas-bearing Medicine Hat and
Milk River formations resulted in signicant production increases, even after the wells had produced
for more than 30 years. Enerplus Resources Fund used both coiled tubing and snubbing-unit tubingconveyed stimulation techniques.

51

52

100,000
Average during the first month for all
12 wells: 6.6 MMcf/D after refracturing
Projected decline
after refracturing

10,000

Total average gas rate, Mcf/D

Short Shut-In Time Well-Test Analysis


Determining how a well should respond to
refracturing requires knowledge about the original fracturing treatment and the current state of
well stimulationfracture length and conductivity. Another objective of the 1998 GTI
restimulation project was to develop a welltesting method to verify restimulation potential
in tight-gas wells.
In low-permeability reservoirs, long shut-in
timessometimes several days, weeks or even
monthsare required to obtain a unique reservoir and fracture characterization from a
pressure-transient well-test analysis, typically a
pressure-buildup test. Consequently, many
operators nd the high costs of performing these
tests and associated production downtime unacceptable. However, if the objective is only to
verify that a well requires stimulation, a unique
well-test solution may not be needed.
Schlumberger developed the short shut-in
time interpretation (SSTI) method to obtain
interpretable well-test data in low-permeability
gas wells.22 This new technique, applicable in
new or depleted reservoirs, uses early-time pressure-transient data to estimate probable ranges
of reservoir permeability and fracture length.
The SSTI method is especially effective in lowpermeability formations, tight-gas reservoirs and
in wells with large wellbore-storage volumes.
This approach is not a quantitative determination of reservoir properties and stimulation
effectiveness, but it is not entirely qualitative
either. The SSTI method defines lower and
upper values for both reservoir permeability
and fracture length at critical points during a
well test. By providing a range of results rather
than multiple sets of nonunique solutions, this
quick and simple determination reduces
uncertainty and nonuniqueness compared with
conventional interpretations.
Reasonably good estimates of reservoir
properties are usually obtained in as little as a
few hours, and generally fewer than three days.
This significantly reduces well-test cost, in
terms of equipment, services and delayed
production. Identifying radial or linear ow into
a well gives a good indication of whether the
current propped fracture is effective or ineffective. The SSTI approach suffers from limitations
in multilayered reservoirs, but engineers can
often use these results to determine if a well
should be restimulated.
The GTI project included a well-testing
program in the Frontier formation of the
North Labarge Unit in Sublette and Lincoln
Counties, Wyoming, USA, to validate restimulation candidates selected by the three GTI

1000
Rate for all 12 wells: 1.5
MMcf/D before refracturing
Projected decline had the
wells not been refractured
100

12 wells refractured at time 0


10
-96

-84

-72

-60

-48

-36

-24

-12

12

24

36

48

60

Normalized time, months

> Kerr-McGee South Texas refracturing results.

methodsproduction statistics, pattern recognition and type curves. The SSTI method was
applied to determine initial hydraulic fracturing
treatment effectiveness in wells at this test site.
Successful application in several Frontier area
gas wells demonstrated the potential of the SSTI
method, but data quality and acquisition
difficulties hampered complete analysis of the
well-test data.
Interpretations using the SSTI method
require high-quality, precise data. Downhole
measurements with precise electronic gauges
and frequent data sampling help capture the
required level of detail. Downhole shut-in
devices reduce wellbore storage effects and
accelerate the onset of linear flow. Using test
times that fall between the start and end of
linear ow, the SSTI method is also applicable in
conventional well tests.

Production-Enhancement Evaluation
Kerr-McGee Corporation and Schlumberger
began working collaboratively to enhance
production from mature, or browneld, South
Texas gas properties in March 2002. These
efforts are the result of a comprehensive reservoir evaluation performed by Schlumberger to
develop a better understanding of completion
and production trends in the Vicksburg basin.
Initiated in the fall of 2001, this proactive study
concentrated on areas where application of new
technologies and techniques would have the
most impact and, in turn, help operators
produce gas more economically.
The objective was to understand how geological, petrophysical and well-completion practices
impact well performance. This Vicksburg study
identified underperforming wells and specific
technologies, such as advanced formationevaluation tools, improved well-completion

22. Bastian P: Short Shut-in Well Test Analysis for


Verifying Restimulation Potential, presented at the
GRI/Restimulation Workshop, Denver, Colorado, USA,
March 15, 1999.
Huang H, Bastian PA and Hopkins CW: A New Short
Shut-In Time Testing Method for Determining Stimulation
Effectiveness in Low Permeability Gas Reservoirs,
Topical Report, Contract No. 5097-210-4090, Gas
Research Institute, Chicago, Illinois, USA
(November 2000).
23. Bradley HB: Petroleum Engineering Handbook.
Richardson, Texas, USA: Society of Petroleum Engineers
(1992): 55-155-12.
Economides MJ and Nolte KG: Reservoir Stimulation,
Third Edition, West Sussex, England: John Wiley & Sons
Ltd. (2000): 5-15-28.

Duda JR, Boyer II CM, Delozier D, Merriam GR,


Frantz Jr JH and Zuber MD: Hydraulic Fracturing: The
Forgotten Key to Natural Gas Supply, paper SPE 75712,
presented at the SPE Gas Technology Symposium,
Calgary, Alberta, Canada, April 30May 2, 2002.
24. Pospisil et al, reference 3.
Olson, reference 3.
Wright and Conant, reference 11.
Marquardt MB, van Batenburg D and Belhaouas R:
Production Gains from Re-Fracturing Treatments in
Hassi Messaoud, Algeria, paper SPE 65186, presented
at the SPE European Petroleum Conference, Paris,
France, October 2425, 2000.
25. Oberwinkler C and Economides MJ: The Denitive
Identication of Candidate Wells for Refracturing,
paper SPE 84211, presented at the SPE Annual Technical
Conference and Exhibition, Denver, Colorado, USA,
October 58, 2003.

Oileld Review

practices and restimulation techniques, which


could have the most impact on well productivity.
The study team gathered and interpreted key
information, including well logs and data related
to fracture stimulation practices. They then
combined elements of both routine and
advanced proprietary processes into an integrated workflow that identified numerous
refracture opportunities.
Key elements of this workflow included the
Moving Domain technique for rapid assessment
of producing properties, development of
a specific petrophysical model to identify
bypassed gas zones and techniques to assess and
mitigate risk. Moving Domain analysis provides a
statistically based analysis of production data to
identify areas with potential for infill development, recompletion and restimulation.
As a result of the project teams efforts,
Kerr-McGee refractured 12 wells during 2002.
Initially, this refracturing campaign added
5.5 Bcf [157.5 million m3] of incremental recoverable gas reserves (previous page). This
equates to $600,000 of revenue per month at
$4/Mcf gas, which increased Kerr-McGee gross
cash ow by an estimated $8.5 million in 2002.
To date, the program has been even more successful in 2003 with an additional 3.6 Bcf
[103.1 million m3] of recoverable gas for the rst
four wells alone. From 2002 to 2003, development costs also were reduced by more than 40%
through improved risk assessment and mitigation.
Schlumberger works with Kerr-McGee across
several geographic locations to facilitate project
execution. Results from the work performed on
each well are published in an Informed Decision
Report (IDR) that includes reservoir properties
derived from the Vicksburg-specific petrophysical model, FracCADE fracturing design and
analysis software, ProCADE well analysis software rate predictions and key production
characteristics from the Moving Domain analysis.
These results are then posted electronically
using InterACT real-time monitoring and data
delivery and made available to Schlumberger and
Kerr-McGee staff participating in the project.
Current teleconference capabilities and collaboration tools, such as InterACT software that
allows review and evaluation of project results as
they become available, facilitate this interaction
and collaboration by the project team.
A Schlumberger project manager located in
the Kerr-McGee office coordinates operations
that range from initial diagnostic work
pressure-buildup tests and production logsto
actual refracturing designs, execution, real-time
monitoring and post-treatment evaluations.

Autumn 2003

160
140
120
100
80

Drill and complete


Coiled tubing
restimulation
Snubbing-unit
restimulation

60
40
20
0

Average cost per well,


$1000

Average production
increase, Mcf/D

> Refracturing economics. In shallow-gas wells like those in the Medicine


Hat and Milk River formations of southeastern Alberta, Canada, restimulating
existing wells costs less (left) and provides incremental production at a lower
unit cost (right) than drilling and completing new wells. To some greater
degree, the same holds true for currently producing wells in many other
elds, especially those in deeper low-permeability reservoirs.

Recognizing the value of a collaborative relationship with Schlumberger, including an extended


staff of experts for the duration of a project,
Kerr-McGee recently identied other browneld
opportunities for joint evaluation.
Restimulations Work
With world demand for petroleum growing daily,
well restimulations are increasingly important.
High productivity improvements for a relatively
low investment make hydraulic fracturing,
either as initial treatments or restimulations,
one of the most economically attractive
production-enhancement techniques.23
Fracture stimulation during initial completion or later in the life of a well bypasses
near-wellbore damage and increases connectivity with the reservoir. The practice of
refracturing began soon after the introduction of
hydraulic fracturing in about 1947, but early
applications required considerable effort to
diagnose problems and select well candidates,
and yielded mixed results. From the 1996 and
1998 GTI studies and associated field trials to
continuing restimulation success in North
America and other areas, including China,
Algeria, Brazil and Russia, it is clear that signicant refracturing potential exists worldwide,
even in mature oil elds.24
In many cases, refracturing is much less
expensive than a new development well and can
inexpensively supplement infill drilling, especially in deep, low-permeability reservoirs. This
is clearly evident even in the shallow-gas wells of
Canada (above). However, restimulation economics are most sensitive to proper candidate
selection. Relatively minor miscalculations can

turn a potentially profitable project into an


unsuccessful venture.
Basically, refracturing candidates are selected
the same way as initial fracturing candidates,
except there may be considerably more data to
work with. Several emerging methodologies,
including multidimensional crossplots and
self-organizing maps, offer operators large
databases that contain hundreds of different
wells, input parameters and fracturing criteria.
In general, these techniques fall into the category of data mining and knowledge discovery.25
Schlumberger also continues to develop and
refine methods for selecting fracturing candidates. Using Moving Domain analysis, for
example, Schlumberger is evaluating ways to use
offset-well production histories as a means of
selecting high-potential refracturing candidates.
When applied judiciously, refracturing has
proved effective for capturing incremental
reserves and the financial benefits they
represent, particularly in todays challenging
business climate. These types of well restimulations are a viable and economically attractive
means of improving economic returns for operators willing to apply new methods and
technologies related to candidate selection and
treatment design.
MET

53

You might also like