Unconventional Gas Reservoir Evaluation. What Do We Have to Consider - Clarkson, 2012

8/17/2019 Unconventional Gas Reservoir Evaluation. What Do We Have to Consider - Clarkson, 2012

1/25

Unconventional gas reservoir evaluation: What do we have to consider?q

C.R. Clarkson a,*, J.L. Jensen b, S. Chippereld c

a Department of Geoscience, University of Calgary, 2500 University Drive NW Calgary, Alberta T2N 1N4, Canadab Department of Chemical and Petroleum Engineering, University of Calgary, 2500 University Drive NW Calgary, Alberta T2N 1N4, Canadac Santos Limited, 60 Flinders Street, Adelaide, South Australia 5000, Australia

a r t i c l e i n f o

Article history:

Received 30 November 2011Accepted 19 January 2012Available online 2 March 2012

Keywords:

Unconventional gasUGR evaluationUGR sample analysisUGR PetrophysicsUGR rate- and pressure-transient analysis

a b s t r a c t

There has been a rapid evolution of technology used to evaluate unconventional gas reservoir andhydraulic-fracture properties, and there currently are few standardized procedures to be used as guid-ance. Therefore, more than ever, petroleum engineers and geoscientists are required to question datasources and have an intimate knowledge of evaluation procedures.

We propose a workow for the optimization of unconventional gas reservoir (UGR) eld developmentto guide discussion of UGR evaluation. Critical issues related to reservoir sample and log analysis, rate-transient and production data analysis are raised. Further, we have provided illustrations of each step of the reservoir evaluation process using tight gas examples. Our intent is to provide some guidance for bestpractices. In addition to reviewing existing methods for reservoir evaluation, we introduce new methodsfor measuring pore size distribution (small-angle neutron scattering), evaluating core-scale heteroge-neity, log-core calibration, and evaluating core/log data trends to assist with scale-up of core data. Ourfocus in this manuscript is on tight and shale gas reservoirs.

2012 Elsevier B.V. All rights reserved.

1. Introduction

Reservoir evaluation involves the determination of reservoirproperties from reservoir samples (ex. core), logs and pressure- andrate-transient data. For conventional reservoirs, there are “triedand true” methods for evaluation that are an outcome of relativelywell understood uid storage and transport mechanisms for thesereservoir types. Many techniques for quantifying key reservoirproperties controlling storage and ow, calculating hydrocarbonsin place, establishing recovery and forecasting production havea long history of development and renement. The reality forunconventional gas reservoirs (UGRs), including low permeability(tight gas), coalbed methane (CBM) and shale gas reservoirs, is thatuid storage and transport mechanisms are poorly understood, and

we are at an early stage for some reservoir types (ex. shale gas) inthe development of such methods. Further, it is not suf cient tocharacterize only the reservoir in unconventional plays but also theinduced hydraulic fracture(s) or fracture network that have a largeimpact on well performance, yet methods for evaluating hydraulicfracture properties are also in their infancy. Indeed there are newmethods for unconventional reservoir and hydraulic fracture

analysis (ex. microseismic monitoring and analysis) that areconsidered critical to the evaluation process that have onlyroutinely been used for oileld applications in the past decade; it isthe job of the UGR evaluator to keep on top of new developments,understand the uncertainties and the consequent impact on theirevaluations.

Perhaps more than any other technical discipline related topetroleum geosciences and engineering, the reservoir engineer isrequired to interrogate, integrate and assimilate data from a vastarray of sources. One only need consider the inputs required toperform a eld-scale reservoir simulation to see that this is true.Inputs include uid, rock and reservoir properties, structural infor-mation about the reservoir, hydraulic fracture properties (for indi-vidual well performance modeling), wellbore architecture and

current and historical completion (event sequences) information,production data and owing pressures, wellbore tubulars andproduction strings, surface hydraulic networks and associatedconstraints. All data sources need to be intensely scrutinized to yieldmeaningful results. The reservoir engineer is usually involved in thedesign of datagatheringprograms (ex. a surveillance program), so anawareness of cost and value of information is also necessary. Finally,the conscientious reservoir engineer will need to evaluate allmethods for arrivingat critical inputs (ex. permeabilityand porosity),and understand sources of uncertainty. Because uncertainty in keyreservoir and hydraulic fracture properties for UGRs is often great,this aspect of the reservoir engineer’s role is particularly critical.

q 2011 Society of Petroleum Engineers. Reproduced with permission of thecopyright owner. Further reproduction prohibited without permission.* Corresponding author. Tel.: þ1 403 220 6445.

E-mail address: [email protected] (C.R. Clarkson).

Contents lists available at SciVerse ScienceDirect

Journal of Natural Gas Science and Engineering

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c om / l o c a t e / j n g s e

1875-5100/$ e see front matter 2012 Elsevier B.V. All rights reserved.

doi:10.1016/j.jngse.2012.01.001

Journal of Natural Gas Science and Engineering 8 (2012) 9e33

mailto:[email protected]://www.sciencedirect.com/science/journal/18755100http://www.elsevier.com/locate/jngsehttp://dx.doi.org/10.1016/j.jngse.2012.01.001http://dx.doi.org/10.1016/j.jngse.2012.01.001http://dx.doi.org/10.1016/j.jngse.2012.01.001http://dx.doi.org/10.1016/j.jngse.2012.01.001http://dx.doi.org/10.1016/j.jngse.2012.01.001http://dx.doi.org/10.1016/j.jngse.2012.01.001http://www.elsevier.com/locate/jngsehttp://www.sciencedirect.com/science/journal/18755100mailto:[email protected]


2/25

The purpose of this manuscript is to discuss evaluationmethods specic to UGRs with an emphasis on understandingdata sources and data uncertainty. To guide this discussion, wepropose a workow to demonstrate one approach tooptimizing eld development (Fig. 1). The evaluation processstarts at small scales (core or ner) and progresses to largerscales (Fig. 1).

Typically, eld development optimization involves a reservoirand hydraulic fracture characterization step, where key uid,reservoir (Table 1) and hydraulic fracture properties (Table 2) areestimated and integrated for the purpose of providing input toreservoir and hydraulic fracture models. This stage involves notonly data collection and property estimation, but also criticaldecisions on the types of models needed to describe the reservoir,as will be discussed below.

Data for different scales are involved in the characterization stepe it is best to consider these scales in the context of uid transportpathways from reservoir to wellbore (Fig. 2). Each characterizationmethod is designed to sample different reservoirsizes and estimatedifferent properties. For example, well-test and rate-transientanalysis are designed to sample meso- (10s of meters) andmacro-scale properties (100s of meters) of the reservoir, and

averaging of important properties (ex. permeability) is performedat this scale. All other methods represent progressively nerreservoir sizes.

We now discuss the reservoir evaluation step in the elddevelopment optimization workow (Fig. 1) for UGRs and addresssome of the primary reservoir engineering considerations. Thereservoir and hydraulic fracture modeling, economic analysis anddevelopment planning steps were discussed in Clarkson et al.(2011a). In some cases, we will also discuss new techniques thatmay prove useful for future UGR evaluation. This review is notmeant to be exhaustive, but rather serves to highlight some of thekey issues related to UGR evaluation. We illustrate the steps andtechniques using tight gas reservoir examples from the sameformation, but different geographic regions (Areas A and B) in

Western Canada. Much of the analysis provided for Area B issummarized from Clarkson et al. (2012). The reader is referred tothat work for further details of analysis procedures, assumptionsand results.

Rate-Transient Analysis

Reservoir Sample and Log Analysis

Additional Data: f luid

properties, f lowing pressures, seismic,

microseismic, spinner &

tracer logs, completions,

wellbore diagrams,

deviation surveys

Reservoir and

Hydraulic FractureCharacterization

Reservoir Simulation Hydraulic Fracture

Modeling

Existing well (vertical

and hz) kh, x f , F cd , Acm

Economic Model

x f or Acm versus

tonnage, volumes, pumping schedule

History match existing

wells, forecasts forvarious scenarios

Economic indicators for

various scenarios

Optimal Design

Absolute permeability, porosity, pore

size distributions, capillary pressure and

relative permeability, electrical

properties, fluid saturations, organic

matter content, gas content, adsorption

isotherms, rock mechanical properties

Modeling

Economic Analysis

Development Planning

Pre- and Post-Frac

Welltest Analysis

Existing well (vertical

and hz) kh, x f , F cD, Acm

Fig. 1. Illustration of a work

ow used to optimize

eld development in unconventional gas reservoirs.

Table 1

Common sources for key UGR reservoir, uid and rock properties. Modied fromSondergeld et al. (2010a).

Reservoir Property Data Source

Porosity Helium gas expansion, mercury injection capillarypressure (MICP), nuclear magnetic resonance(NMR), log analysis (calibrated to core)

Permeability Core Analysis: Steady-state and unsteady state

(pressure- and pulse-decay), MICPWell-test Analysis: (pre- and post-fracture)Injection/falloff (IFOT), diagnostic fracture injectiontest (DFIT), post-fracture ow and buildup (FBU)Production Analysis: Rate-transient analysis (RTA),simulation history-matching

Pore Pressure IFOT, static pressure/temperature gradient,perforation inow diagnostic (PID), perforationinow test analysis (PITA), various openhole tests

Water saturation Core extraction (Dean Stark, Retort), capillarypressure, log analysis (using lab-based electricalproperty measurements)

Free and sorbed gas Desorption canister testing & adsorption isotherms,calibrated log analysis

Total Organic Carbon Leco TOC & RockEval (calculated)Thermal Maturity Vitrinite reectance (Ro), RockEval (calc)Rock composition X-Ray Diffraction (XRD),Fourier-Transform Infrared

(FTIR), visual point count (for optically-resolvablegrains), EDAS (SEM), electron microprobe

Rock mechanicalproperties

Core measurements, log-derived (multi-componentsonic, ex. DSI)

Fracture andclosure stress

Mini-frac tests, DFIT, log-based (DSI) calibrated tocore

Fluid properties Mud log, produced gas and water analysis (PVTproperties)

Temperature Openhole logs, static pressure/temperaturegradient, production logs

C.R. Clarkson et al. / Journal of Natural Gas Science and Engineering 8 (2012) 9 e 3310


3/25

2. Reservoir sample analysis

Reservoir sample analysis includes the evaluation and integra-tion of a variety of measurements, including full-diameter core,core plugs, cuttings and analog outcrops, to assess properties suchas absolute permeability, porosity and sorption characteristics(Table 1). New methods have emerged for UGR evaluation. Thereare often few standards developed for evaluating key reservoirproperties, and the reservoir engineer must be aware of the vari-ability in estimated properties depending upon sample capture,storage and preparation techniques, and analytical method used. Asalways, data measured on different scales (ex. Fig. 2) have to beintegrated to understand the reservoir behavior.

Table 1 provides an overview of the techniques currently usedfor UGRs, which will be briey reviewed below. First we begin witha discussion of the importance of sample collection, handling/preparation and type.

2.1. Sample handling and type

Sample conditions include the state of the sample (ex.preserved or open to the atmosphere), its size and volume, the

mud characteristics and their impact on measured properties,sample representativity of the interval of interest, and the abilityto reproduce in-situ conditions (ex. stress and uid saturation).Lack of sample preservation often results in uid loss or gain andgas ow in UGRs is extremely sensitive to water saturation (S w), asis gas sorption on organic matter. Standardized tests (ASTM) donot exist to reproduce in-situ water content in tight gas or shale.Fig. 3 illustrates the relationship of sample size and dehydrationtime for a tight gas reservoir. Both samples were subject toatmospheric pressure and 20 C; short exposure of the crushedsample caused an S w decrease from 40% to 2%, while the core plugS w decreased after a much longer period. Clearly sample sizeaffects the dehydration rate, and even modest exposure can affectS w estimates. Thus, preserved samples are recommended foranalysis.

Sample size affects the volume over which petrophysical prop-erties are averaged (ex. Corbett et al., 1998). Whole (full-diameter)core may be more representative than smaller samples, particularlyif ne-scale heterogeneities are present such as laminations, butwhole core may be more susceptible to drilling mud invasion,altering properties and being more dif cult to clean. Core plugsandsidewall cores may provide more precise estimates of properties of

specic lithofacies, sampled at a ner scale than full-diameter core,and may improve poro-perm correlations, but there is a possibilityof sampling bias. On viewing core, core plugs are often taken fromintervals that appear to be the best reservoir and such practicesplace a bias on the sampling program, causing problems with log-core analysis and perhaps missing impermeable elements that maycontrol production. Care must be taken in cutting core plugs whenswelling clays are present; for this purpose, either brine, or pref-erably liquid nitrogen should be used as a lubricant. Percussionsidewall cores can lead to loss of sample integrity and fracturing,and cuttings are problematic due to possible contamination withnon-reservoir lithologies, and dif culty in accurate determinationof the original sample depth. Mavor (1996) discusses samplingprotocols for coalbed methane to improve core-log calibration for

gas content determination, some of which are applicable toorganic-rich shales.

There is a fundamentally important philosophical differencebetween commercial labs as to whether it is betterto crush samplesbefore permeability and porosity measurement, or to perform the

Table 2

Common sources for hydraulic fracture properties for UGRs.

Hydraulic Fracture Properties Data Source

Propp’d hydraulic fracturelength and conductivity (static)

Post-fracture net-pressure analysis(hydraulic fracture modeling),post-fracture ow and buildup

Propp’d hydraulic fracturelength and conductivity (owing)

Rate-transient analysis

Created hydraulic fracturelength, height and geometry(complexity)

Microseismic, tiltmeter survey,4D seismic

Macrofabric

Modified from Bustin, Bustin and Cu i, 2008

Macroscale (reservoir)

Mesoscale

(microfracture

network)

Microscale (nanopore

network)

Nanoscale (gas

desorption from

nanopore walls)

Molecular (mass

transfer from

kerogen/clay bulk

to pore surface)

Fig. 2. Illustration of the impact of scale on transport mechanisms in shale gasreservoirs. Flow to the wellbore is rst initiated at the macro-scale, followed by ow atprogressively ner scales, including molecular transport through nanoporosity in

kerogen. Modi

ed from Javadpour et al. (2007).

Fig. 3. Illustration of the effect of sample size on dehydration time for a tight gas

sample from study area A.

C.R. Clarkson et al. / Journal of Natural Gas Science and Engineering 8 (2012) 9 e 33 11


4/25

measurements on core plugs or full-diameter core; the differencesin results may be quite signicant (Sondergeld et al., 2010a). Thereare advantages and disadvantages to both, but we note that thereare variable procedures between labs regardless of what samplesizes are used. The process of reservoir sampling often causesstress-relief fracturing to occur in the samples. Some labs prefer tocrush samples to remove these microfractures that were notpresent in-situ and to provide access to more non-connectedporosity. There are several disadvantages to this method,including the inability to provide stress to crushed samples and thefact that permeability is a function of particle size (Cui et al., 2009).Core plugs sample a larger volume of reservoir than crushedsamples, and hence measurement involves averaging over a largerscale. If ne laminations or fractures occur in the samples, thenorientation of the plug relative to the heterogeneities has a largeimpact on the permeability. With core plugs, measurements tend totake longer, and stress-relief fractures may be present. Measure-ments at elevated conning pressure may close the articially-induced fractures.

A critical issue for UGRs is the recreation of in-situ conditions inthe lab. Byrnes (2006) notes that all measured petrophysical prop-erties are a function of the pressure (pore and conning pressure),

volume/scale, temperature and time/history, and composition (rockand uid) (PVTxt). Lab measurement conditions directly inuencethe property being measured. For example, permeability andporosity are functions of the pore/conning pressure and method forapplication of stress (uniaxial, hydrostatic conditions etc.), currenttemperature of measurement, temperature and pressure history, anduid type used. The reservoir engineer must always question the labmeasurement conditions and understand the sensitivity of thoseproperties being measured to PVTxt.

We now discuss specic measurements performed during thereservoir sample analysis step (Fig. 1) and issues relating to themeasurement that the reservoir engineer should know. Examplesillustrate the issues involved.

2.2. Matrix uid saturation and moisture content

Fluid saturation and moisture content affect free gas and sorbedgas storage (coals, some shales) in UGRs. Common methods forsaturation measurement in tight gas and shale reservoirs are DeanStark extraction and retort; crushed samples are often used forextraction from shale. Both Dean Stark and retort methods havebeen standardized (API e RP40, API, 1998). These API standards areapplicable to conventional and tight reservoirs with permeabilitiesdown to 1 md (Sondergeld et al., 2010a).

A principle source of error with retort and Dean Stark methodsis the saturation change experienced by the core during coringoperations, retrieval and preparation. Sondergeld et al. (2010a)compared shale gas reservoir water saturations, S ws, obtained

from commercial labs using retort and Dean Stark methods,demonstrating substantial differences. They noted that there is nostandardization of measurements for shale gas reservoirs.

For these extraction methods, sample preservation is critical(Fig. 3). In one well in tight gas study Area A, three 36 m cores weretaken and subject to routine porosity, permeability and Dean Starkextraction (using toluene). Over 100 1.5” diameter core plugs werecut from the preserved cores using liquid nitrogen as a lubricant.S ws ranged from 10.5 to 99.3%, oil saturations from 1.1 to 25.1% andgas saturations from 0 to 84.5%. S ws were generally 200 ms). NMR was recently used to study shalewettability (Odusina et al., 2011). We expect this technology willplay an increasing role in UGR uid saturation and pore size anal-ysis in the future.

2.3. Matrix porosity and pore size distribution

Porosity and porosity distribution estimates are critical for bothstorage and ow quantication in UGRs. Due to the complex natureof UGR porosity, the typically wide distribution of pore sizes (PSD)and shapes, and nanometer pore sizes, measurement of porosityand PSDs can be a complicated affair (Fig. 4). Both uid invasionmethods, where a molecular probe is used to investigate the porestructure (connected pores only), and radiation methods, where x-ray or neutron scattering patterns are interpreted in terms of porestructure, have been used (Bustin et al., 2008a). If properly per-formed, porosity measurements (combined with saturationmeasurements) are useful for estimating free-gas storage, and PSDscan be used to assess storage mechanisms, infer transport mecha-nisms, and assist with establishing controls on uid ow (Fig. 5).

The most common methods for estimating porosity and PSDs inshales and tight gas reservoirs remains helium pycnometry forgrain density (combined with a measurement of bulk density) toestimate connected porosity, and mercury injection (MICP) toobtain a PSD. As noted by Bustin et al. (2008a), use of helium maylead to errors in effective porosity estimation because the heliumkinetic diameter is smaller than most reservoir gases, so a greaterpercentage of total porosity will be investigated than is accessiblewith reservoir gases. If reservoir gases are used, corrections foradsorption must be made. Crushed versus whole core or core plugscould lead to different results for porosity, and because porosity isstress-sensitive (particularly for those samples that containmicrofractures or slot porosity), it is best to measure porosity on

100,000

10,000

1,000

100

10

1

Mesopores

2 – 50 nm

Micropores< 2 nm

Macropores> 50 nm

IUPACClass.

Analysis Methods

O p t i c a l M i c r o s c o p y

S E M

T E M

S A N S / U S A N S

Pore Diameter nm

Radiation Penetration-Fluids

M

e r c u r y P o r o s i m e t r y

N i t r o g e n

C a r b o n D i o x i d e H

e l i u m P

y c n o m e t r y / M e r c u r y D e n s i t y

Fig. 4. Methods used to estimate porosity and pore size distributions in unconven-

tional gas reservoirs. Modi

ed from Bustin et al. (2008b).



5/25

core samples subject to conning stress. Mercury porosimetrysuffers as a measure for pore size for several reasons. It is limited topore sizes generally >2 nm (and hence excludes micropores inorganic-rich shale, which could contribute signicantly to matrixporosity), is likely to distort the pore structure of highly-compressible samples because of the high-injection pressuresrequired for measurement (up to 60,000 psi), and requires esti-mates of surface tension/contact angle for pore size calculationsfrom the Washburn (1921) equation, which are not accuratelyknown for UGRs (and may vary by sample type). A furtherconsideration for porosity and PSD measurement is that UGRs arevery heterogeneous, and hence a large number of samples may berequired to statistically characterize the reservoir e often sample

sizes are greater than the scale of the measurement necessitatingsome averaging to occur.Bustin et al. (2008a) noted that low-pressure adsorption tech-

niques can also be used to characterize pore structure of shales.Low-pressure nitrogen adsorption isotherms can be interpreted forpore size distribution in the mesopore range (pore diametersbetween 2 and 50 nm), but have an upper pore size limitation of around 300 nm. Micropore (pore diameter < 2 nm) volume can beestimated from CO2 adsorption.

Radiation methods include scanning electron microscopy,transmission electron microscopy, small-angle x-ray and neutronscattering and optical microscopy. Optical microscopy allows onlythe largest pores in unconventional reservoirs to be observed, butis very useful for understanding of mode of occurrence and

porosity types (ex. intergranular, intraparticle, slot porosity),

mineralogic controls and affects of diagenesis (mechanical andchemical). Pores that can be observed with optical microscopy,assuming they are open at in-situ conditions, affect uid ow andstorage. Recently, backscattered scanning electron imaging,combined with focused ion beam (FIB) milling techniques, haveallowed observations of porosity to be extended down to thelower mesopore range, and have allowed porosity associations tobe interpreted at this scale. Although occurrence of nanoporosityin kerogen has been known for many years (ex. Gan et al., 1972),we have only recently been able to identify pore sizes to w4 nmwith SEM. Some have even used the technique to create 3Dvolumes of porosity (Sondergeld et al., 2010b), but we note thatthese volumes are usually at the micron scale and would be

dif cult to scale-up for engineering purposes. Scanning trans-mission electron microscopy (STEM) methods have also recentlybeen applied to shales, which allow even smaller pore sizes to beresolved, but scale-up of sample volume is challenging with thistechnique.

Not included in Fig. 4, some researchers have also investigatedthe use of nuclear magnetic resonance (NMR) as a way to investi-gate pore structure in coal and shale. Recently, Sondergeld et al.(2010b) presented T 2 spectra for Barnett shales and noted that“NMR is sensing pore bodies not accessed by mercury data” andcalculated pore body radii from 5 nm to 150 nm.

Very recently, small-angle neutron (SANS) and ultra-small-angle neutron (USANS) measurements have been used on coalbedmethane reservoir samples to investigate matrix pore size distri-

butions and the impact of CO2 adsorption on pore structure

0.00001

0.0001

0.001

0.01

0.1

1

10

100

1000

10000

100000

a b

c

1 . 0 E -1 1

1 . 0 E -1 0

1 . 0 E - 0 9

1 . 0 E - 0 8

1 . 0 E - 0 7

1 . 0 E - 0 6

1 . 0 E - 0 5

1 . 0 E - 0 4

1 . 0 E - 0 3

1 . 0 E - 0 2

Pore Diameter (meters)

R a t i o o f F r e e

G a s t o A d s o r b e d G a s

“Normal” porosity

Methane molecule 0.4 nm

1 µm

Microporosity

1 nm

1E-06

1E-05

1E-04

1E-03

1E-02

1E-01

1E+00

1E+01

1E+02 1E+03 1E+04 1E+05

K n

Pressure (kPa)

50 µm 10 µm 1 µm

300 nm 50 nm 10 nm

N o - s l i p

f l o w

T r a n s i t i o n

f l o w

0.00001

0.0001

0.001

0.01

0.1

1

0 5 10 15 20 25 30

P e r m e a b i l i t y ( m D

)

Porosity (%)

pore size from N ads

0.2 µm

0.1 µm

0.05 µm

Fig. 5. Uses of pore size distributions for unconventional reservoir characterization: a) inference of gas storage mechanism (modied from Beliveau, 1993); b) inference of transfermechanism (modied from Javadpour et al., 2007); c) and identication of dominant pore throat radius using a Winland-style plot (modied from Clarkson et al. 2012). Lines on (c)represent r p35 (pore throat aperture (mm) at 35% Hg saturation during a mercury injection test) using the Eq. (2) proposed by Aguilera (2002, 2010).



6/25

(Melnichenko et al., 2009). In a SANS experiment (Fig. 6), a neutronbeam is directed at a sample, and the neutrons are elasticallyscattered due to their interaction with nuclei of atoms in the samplee the scattering vector is related to a characteristic length scale(pore size) in the sample. The appeal of this method is thatexperiments can be performed at high pressure and temperature(simulating reservoir conditions), allowing for changes in porestructure to be observed as a function of pressure and, undercertain conditions, open versus closed porosity may be distin-guished. SANS experiments, combined with USANS, also enablea wide distribution of pore sizes to be investigated.

2.3.1. Tight gas examples

Our tight gas study areas are used to illustrate the impact of sample heterogeneity and stress on porosity and lithology differ-ences on PSDs. Further, ow mechanisms and storage mechanismscan be inferred, which inuences basic reservoir engineeringanalysis.

In both areas, connected porosity was established using heliumexpansion. PSDs were obtained from mercury intrusion in Area A(43 core plugs), while seldom applied techniques (for tight gas) of low-pressure N2 adsorption and SANS (3 core samples) were

applied for Area B. In Area B, routine porosity measurements wereperformed on 37 full-diameter cores and 25 core plugs. Of the 25core plugs, 10 were selected for measuring porosities underhydrostatic stress (set to represent in-situ conditions) and of those10, 3 were selected for measuring porosities under variableconning pressure. The 3 cores selected for variable conningpressure measurement were sub-sampled for N2 adsorption anal-ysis. The lithologies of the 3 samples differed primarily in theamounts of quartz, dolomite, orthoclase and calcite.

The impact of conning pressure for the three core plugs in AreaB is signicant (Fig. 7). The difference between porosities measuredat “ambient” and net overburden conditions averaged about 11%,with some variability (8e19%) between samples (Fig. 7a), sug-gesting lithological differences between the samples may play

a role (Fig. 7b).Area B PSDs were obtained using nitrogen adsorption/desorp-

tion analysis (Fig. 8). This technique is seldom used for tight gasreservoirs (although more so for shale gas and coal), but proveduseful here. The adsorption isotherms (Fig. 8a) can be interpretedfor PSDs (Fig. 8b) using BJH Theory (Gregg and Sing, 1982; Clarksonand Bustin, 1999). The PSDs for two of the samples are bimodal,with the pore size fraction in the 50e100 nm range (500e1000 Å).From the interpreted size range, non-Darcy (slip ow) can beimportant, particularly at low-pressures, which will inuence themodel choice used for simulation and production analysis (Fig. 5b).Free-gas storage is expected to dominate in this reservoir due to theabsence of microporosity (


7/25

permeability and porosity evaluation will vary by reservoir type,however.


Our tight gas study areas are used to illustrate the impact of stress and compositional heterogeneity on absolute permeabilitymeasurements. The tight gas reservoir in Areas A and B consists of a nely interbedded very ne grained sandstone, siltstone andshale with a strong contrast in permeability from lamination tolamination. In Area B, several different permeability measurementtypes were taken, including routine permeabilities measured ona full-diameter core at 500 psia net conning pressure, pressure-decay probe (prole) measurements on the slabbed core andpulse-decay permeability performed on 1 inch core plugs taken atthe same locations as several of the probe measurements. The

routine permeability values have a lower resolution of 0.01 mD asa time-limit was placed on reaching steady-state, meaning thatpermeability at many of the locations could not be accuratelydetermined (Fig. 11).

Prole permeability data were collected on the slabbed coreusing a PDPK-400 (CoreLab) probe permeameter at approximately2.5 cm (1 inch) intervals. Details of probe permeameter measure-ment procedures and theory are provided in Jones (1994). Probemeasurements were taken approximately parallel to bedding in anattempt to characterize ne-scale heterogeneities in the core(Fig. 12). The analyzed core is nely laminated and, because proledata can be collected at a much ner scale than routine whole core

measurements, the probe data reveal more variability. Above2209 m, cyclic (sinusoidal) changes in permeability appear to occurabout every 2 m. Generally, the measurements show a modestrange in values (slip-corrected permeabilities from w0.001 to0.03 mD), suggesting a fairly narrow range in dominant pore throatradius.

Although the prole values provide a quick way to characterizevertical (and potentially lateral) variation in permeability, themeasurements are performed on unconned core and hence arenot representative of in-situ conditions. To investigate stress-dependence of permeability, the 10 core plugs cut for porositymeasurement (discussed above) were also used for pulse-decaypermeability ( Jones, 1997) measurements under hydrostatic stress(set to represent in-situ conditions), and 3 were selected formeasuring permeability under variable conning pressure.

Comparison of pulse-decay permeability measurements atambient conditions and at in-situ conditions (Fig. 13a) revealsconsiderable stress-dependence, with differences as great as 71%,however there is a much larger difference between probe per-meameter measurements at atmospheric pressure compared to thepulse-decay measurements under stress. Much of the permeabilityloss is within the rst few MPa of conning pressure, possibly dueto the presence of open microfractures caused by stress relief (Fig. 13b). There is also some variation in stress-dependence bysample and hence lithology. The relationship between probe andpulse-decay measurements measured at conning pressure(Fig. 14) can be used to “adjust” the probe measurements to in-situ

0

0.025

0.05

0.075

0.1a b

2195 2200 2205 2210 2215

P o r

o s i t y ( F r a c t i o n )

Depth (m)

Confined Ambient

0.02

0.04

0.06

0.08

0.1

0 10000 20000 30000 40000 50000

P o r o s i t y ( F r a c t i o n )

Net Overburden Pressure (kPa)

Sample A Sample B Sample C

Fig. 7. Impact of conning pressure (hydrostatic) on porosity for tight gas samples from Area B. (a) Comparison of porosities measured at ambient conditions and those at 14466 kPa(2098 psi) lab hydrostatic or 25468 kPa (3694 psi) reservoir conning pressure. “Ambient” conditions refer to measurements performed at 2758 kPa (400 psi) lab hydrostatic netconning pressure or equivalent reservoir net conning pressure of 4856 kPa (704 psi). (b) Inuence of lithology and variable conning pressure on porosity. Modied fromClarkson et al. (2012).

Fig. 8. Nitrogen adsorption/desorption isotherms for 3 tight gas samples collected from core plugs from tight gas Area B (a), and interpreted pore size distributions (b) using BJH

Theory. Note the bimodal pore size distributions for samples A and B, with larger pore sizes in the 50e

100 nm (500e

1000 Å) range. Modi

ed from Clarkson et al. (2012).



8/25

stress conditions e this is useful for core-log calibration, as dis-cussed below.

2.5. Capillary pressure and relative permeability

In tight gas and shale gas reservoirs, capillary pressure curvesare used to quantify saturation distributions which are importantfor free-gas storage calculations. Further, it is well knownthat effective permeability to gas is strongly affected by watersaturation, and that relative permeability curves are useful toestablish the dependence of effective permeability on watersaturation.

Newsham et al. (2004) observed that the combination of lowconnate S ws and high capillary pressures in tight gas sands have

prevented the application of conventional methods for estimationof capillary pressure curves; those authors recommended the use of hybrid methods (combined high-pressure porous plate andcentrifuge methods with the vapor desorption techniques) to covera wide range in S ws. Newsham et al. (2004) also recommend usingmethods that use reservoir-like uids. Some of the ndings of Newsham et al. (2004) mayalso extend to the inorganic frameworkof shale gas reservoirs, but the presence of organic matter willaffect wettability, and cause vapor adsorption when air-brinesystems are used. Some authors (Andrade et al., 2010) suggestthat brine saturation is limited to the inorganic framework, but wesuspect this is organic-matter type-dependent as coal (usuallyType-III organic matter) is known to sorb water. To our knowledge,a detailed study of capillary pressure in shale gas reservoirs is yet tobe published. In shale gas reservoirs, Andrade et al. (2010) suggestthat instantaneous capillary equilibrium is not realistic, and thatnon-equilibrium formulations should be considered for simulation.

Relative permeability is similarly dif cult to measure in UGRs.Steady-state and non steady-state techniques have been applied,with associated limitations. Usually, large differential pressures are

Fig. 9. Example pore size distribution for tight gas sample in Area A using MICP. This sample has routine air permeability and porosity values comparable to the 3 samples in Area B.

Note the pore size classication used in the plot on the left is not the same as IUPAC (Sing et al., 1985). Modied from Clarkson et al. (2012).

10-1

100

101

102

103

104

105

106

107

108

109

I n t e n

s i t y [ c m

- 1 ]

10-4

10-3

10-2

10-1

q [A-1

]

Shale#4

Shale#5

Shale#24

USANS SANS

Sample BSample ASample C

Fig. 10. SANS/USANS data from 3 tight gas samples collected from core plugs cut intight gas Area B. q is the scattering vector. Image courtesy of Lilin He and Yuri Mel-

nichenko of Oak Ridge National Labs (2010).

0.01

0.1

0 0.02 0.04 0.06 0.08 0.1

R o u t i n e A i r P e r m e a b i l i t y

( m D )

Routine Porosity (Fraction)

Fig. 11. Crossplot of routine permeability and porosity for 37 full-diameter coresamples in Area B. Twenty-four samples have permeabilities at or below the

measurement lower limit. Modi




9/25

required to initiate 2-phase ow in tight gas reservoirs, creatinglarge saturation and pressure gradients. As with absolute perme-ability, non-Darcy ow effects must be accounted for, and thecorrection is known to vary with saturation and temperature(Rushing et al., 2003). Failure to correct for slippage may causeeffective permeability to gas to be overestimated.

Some studies focus on gathering effective permeability to gasusing the single-phase stationary method (water is held in sampleby capillary pressures while gas ows). Effective permeability togas can then be measured at different S ws e a set of incrementalphase trap experiments is discussed for our tight gas study Area Abelow.


Although tight gas sands often lack mobile water, water may beintroduced to the system through hydraulic fracturing operations.The combination of a water-wet system and high capillary pres-sures causes water to be imbibed changing the effective perme-ability to gas, particularly near the fracture face. The impact of imbibed water on effective permeability to gas was investigated for4 core samples in Area A (Fig. 15) using incremental phase trapexperiments. The initially dried core samples were rst weighed,then subject to a test temperature of 20 C and a net overburdenpressure representative of the reservoir. The initial permeability togas was measured at these (dry) conditions and used as the base-line permeability. Brine was injected to yield S w ¼ 10%, thenhumidied gas was owed in both directions through the core toachieve an even saturation distribution.Saturations were estimatedgravimetrically at each step. Gas permeability using the humidied

gas was measured (at overburden pressure) to establish the impactof the trapped water.

The results suggest that gas permeability is more stronglyaffected by S w for the highest permeability samples compared tothe low permeability samples (Fig. 15). For all samples, perme-abilities are less than 0.0004 md for S ws > 20%, thus imbibition of

2195

2197

2199

2201

2203

2205

2207

2209

2211

2213

2215

0.0001 0.001 0.01 0.1

D e p t h ( m )

Probe Permeability (mD)

Fig. 12. Slip-corrected probe permeability at 593 locations collected on slabbed core intight gas Area B. Modied from Clarkson et al. (2012).

0.0001

0.001

0.01

0.1a b

2195 2200 2205 2210 2215

P e r m e a b i l i t y ( m D )

Depth (m)

Pulse-Decay/confined Pulse-Decay/Ambient Probe

0.0001

0.001

0.01

0.1

0 10000 20000 30000 40000 50000

P e r m e a b i l i t y (

m D )

Net Overburden Pressure (kPa)

Sample A Sample B Sample C

Fig. 13. Impact of conning pressure (hydrostatic) on permeability for tight gas samples from tight gas Area B: (a) Comparison of pulse-decay permeabilities measured at ambientconditions and those at 14466 kPa (2098 psi) lab hydrostatic or 25468 kPa (3694 psi) reservoir conning pressure. “Ambient” conditions refer to measurements performed at2758 kPa (400 psi) lab hydrostatic net conning pressure or equivalent reservoir net conning pressure of 4856 kPa (704 psi). Probe (prole) permeability values measured at the

location of the plug ends are also shown. (b) In

uence of lithology and variable con

ning pressure on permeability. Modi


y = 0.33x0.58

R2

= 0.77

0.0001

0.001

0.01

0.1

10.0100.01000.0

P r o b e P e r m e a b i l i t y ( m D )

Pulse-Decay Permeability (mD)

Fig. 14. Probe vs. pulse-decay permeability measured at net overburden pressure,collected for core in tight gas Area B. The two values are modestly correlated, but differsubstantially in absolute value due to differences in measurement conditions andvolume of rock sampled. Modied from Clarkson et al. (2012).



10/25

water during fracturing operations could have a serious affect ongas permeability.

2.6. Electrical properties

Electrical properties provide a link between pore-scale petro-physical properties and log-derived properties. Conductance isaffected primarily by movement of ions in the pore uids which inturn is affected by pore structure, pore geometry, rock compositionand temperature and hydrocarbon uids. Two parameters, cemen-tation factor (m) and saturation exponent (n) can be derived fromcore-based measurements and are required for S w estimates usingArchie’s equation, although Archie’s equation may not be applicable

for UGRs. Ions in clay-bound water may act as separate conductors,which must be included in S w calculations; the conductance of the

cores is measured using uids of varying salinity and plotted versusthe salinity of the saturation uid. “Pseudo” m and n may be derivedfrom this plot and input into Archie’s equation. m is not just a func-tion of cementation, but is a function of pore geometry too. Aguilera(2008) derived m for tight gas sand reservoirs that contain naturally-fractured and/or slot porosity using a dual porosity model. He notedthat the decrease in m with porosity may be indicative of a slot porenetwork. Recently, Aguilera (2010) noted that the petrophysicalmodel for shale gas reservoirs should include up to four porositysystems: fracture, inorganic matrix, organic matter, and porosityassociated with the induced hydraulic fracture network. Clearly,theoretical m calculations for such systems could be complex.


Electrical properties were measured on Area A samples coveringa range of rock properties. Samples were ushed with 5 porevolumes of brine to ensure S w ¼ 100%, and resistivity readings weremade until they were stable. Formation resistivity factor wasmeasured at several overburden pressures (Fig.16a). The derived mranged from 1.29 to 2.35, suggesting that some samples may havemicrofractures since m approached 1 at low porosities. The dualporosity model of Aguilera (2008) was t to the measured

cementation factors (Fig. 16b); a fracture porosity of 0.5% wasassumed, and the cementation factors for the fractures (m f ) andmatrix (mb) were adjusted to provide a visual best t. The resultssuggest that perhaps that some of the samples may exhibit dualporosity characteristics, but there is much scatter in the data.Petrographic analysis should be performed to ascertain whethersuch a model is warranted.

After the formation resistivity factor measurements, the sampleswere putinto a porous plate desaturation cell, and placedin capillarycontact with the synthetic brine. Humidied nitrogen was used todesaturate the samples at incrementally increasing capillary pres-sures. When capillary equilibrium was reached, the samples wereplaced in a core holder subject to net overburden pressure and theresistivity measured. n varied from 1.6 to 2.21, with a composite

value of 1.87, representing the best t through all sample measure-ments (analogous to Fig. 16a for cementation factor).

0

0.4

0.8

1.2

1.6

2

0 10 20 30 40

G a s P e r m e a b i l i t y ( m i c r o d a r c i e s )

Water Saturation (%)

Core 22 Core 25 Core 34C Core 56B

Fig. 15. Effective gas permeability as a function of S w for 4 Area A tight gas coresamples. Gas permeabilities were measured at the maximum drawdown pressure of 20670 kPa. Modied from Clarkson et al. (2012).

1

10

100

1000

10000

0.001 0.01 0.1 1

F o r m a t i o n R e s i s t i v i t y F a c t o r

Net Overburden Porosity (Fraction)

1

1.5

2

2.5

3

0 0.05 0.1 0.15

C e m e n t a t i o n

F a c t o r , m

Net Overburden Porosity (Fraction)

y = 1.00x-1.83

a b

Fig. 16. Formation Resistivity Factor versus porosity for 18 core samples in tight gas Area A.



11/25

To test the applicability of the m and n parameters derived fromArea A core to other areas, we used Archie’s equation to estimateS ws in Area B (Fig. 17). Although the macroscopic lithology for bothareas appears similar, there are likely differences in mineralogy andpore structure. Using m ¼ 1.83 and n ¼ 1.87 obtained from Area Adoes not yield a good match (red curve) to the Area B core data; thematch (green curve) using best-t values m ¼ 1.38 and n ¼ 1.81 isreasonable and within the ranges obtained from Area A data. Wenote, however, that the core-derived S w may be toolow because thesamples were not preserved. This exercise shows the sensitivity of log-calculated S w values to m, n values, and the danger of applying

composite values from one UGR area to another.

2.7. Gas content and adsorption isotherms

Organic-rich shale plays have a signicant amount of gas storedin the sorbed state, requiring special measurements in the eld/labto quantify. Much of the historical research on sorbed gas storagehas focused on coalbed methane reservoirs; a recent summary of CBM gas content and sorption isotherm measurements, along witha discussion of uncertainties, was provided by Clarkson and Bustin(2010). Unlike most CBM plays, however, free-gas storage is oftensignicant in shale gas reservoirs, and must be included in gas-in-place calculations (Ambrose et al., 2010). Whether or not sorptionaffects gas production from shales depends on the amount of

kerogen in the system and whether it is connected to a fracturesystem (natural or induced).There are many controls on sorbed gas storage: 1) pore pres-

sure; 2) temperature (sorption decreases with increase in temper-ature); 3) moisture content (adsorption decreases with moisturecontent up to a moisture limit); 4) organic matter content (micro-porous structure in organic matter tends to drive the amount of sorption that occurs); 5) organic matter composition (differentorganic matter constituents referred to as macerals have differentpore structures); 6) organic matter thermal maturity (OM porestructure changes with thermal maturity); and 7) gas composition(different gases have different sorption characteristics). Further,Hartman et al. (2008) noted that some clays may contribute tosorbed gas storage in shales and that for oil or wet-gas prone shales,

entrained bitumen may affect sorption of gas. There has been much

focus on quantifying total organic carbon (TOC) in the lab for shales,correlating lab-based sorption measurements to TOC, then usinglogs to predict TOC using a variety of methods (see below), but wesuggest that there may be considerable error in this process if TOCis not the only driver for sorbed gas content. Nonetheless, TOCmeasurements are routinely performed for shales along with anestimate of thermal maturity, usually done using vitrinite reec-tance (Ro) measurements.

Ambrose et al. (2010) noted potential problems with shale gas-in-place calculations if sorbed gas volume is not accounted for infree-gas calculations associated with nanoporosity. Their investi-gation was performed assuming single-component gases only, butHartman et al. (2011) also investigated the impact of sorbed gasvolume if multi-components are present. Hartman et al. (2008)noted that oil- or wet-gas prone shales, which are being activelyexplored for, are complex with respect to uid adsorption andrecommended studies be performed in this area. The following isbrief summary of mixed gas/uid adsorption modeling performedas part of the Hartman et al. (2011) study.

2.7.1. Shale gas example

Hartman et al. (2011) noted that many shale gas plays exhibit

a transition from drygas to liquid-prone areas and further suggestedthat multi-component sorption modeling is particularly critical inthe liquids-rich environment. Historically with shale gas and CBMplays, the Extended Langmuir model has been the most frequentlyapplied for multi-component adsorption modeling, but the limita-tions of this model are well-known (see Clarkson and Bustin, 2000).Hartman et al. (2011) suggested that more rigorous adsorptionmodels should be utilized and compared predictions of binary gasadsorption using the Ideal Adsorbed Solution (IAS) model and theExtended Langmuir (EL) Model. The details of the models are dis-cussed elsewhere (ex. Clarkson and Bustin, 2000); the following isa summary of some of the ndings of Hartman et al. (2011).

Using single component isotherm data methane, ethane,propane, butaneþ and carbon dioxide given in Ambrose et al. (2011),

the predictions of the IAS and EL modelsfor binary mixtures of thesecomponents are presented in Fig.18 in theform of phase equilibriumplots. In all cases, methane is included as the primary component,and binary mixtures of methaneeethane, methaneecarbon dioxide,and methaneebutaneþ and methaneepropane are modeled usingthe two approaches. Predictions are made at two different totalpressures (4000 and 1000 psia) for the IAS model. The predictions of the two models for all mixtures are signicantly different, except forthe methaneþ ethane system at 1000 psia. In all cases, the EL modelpredicts more methane (weakly sorbing component) in sorbedphase than the IAS model. The EL model predicts that the compo-sitions do not change with pressure, which Hartman et al. (2011)believe to be inaccurate.

From the Hartman et al. (2011) study, it is clear that: 1) different

components sorb to shale to varying degrees and if mixtures existin the produced uids, then multi-component adsorption must beaccounted for and 2) the amount of mixed-uid adsorption pre-dicted depends strongly on the choice of the adsorptionmodel used.

2.8. Rock mechanical properties

A comprehensive discussion of mechanical properties of shalesand anisotropy was provided by Sondergeld et al. (2010a). Severalpoints raised by those authors are that shales are elasticallyanisotropic (requiring the measurement of up to 7 elastic constantsif the samples are fractured normal to bedding); sample size mustbe appropriately selected; obtaining analysis on vertical, horizontal

and 45

core plugs can be challenging for shale and there may be

0

0.1

0.2

0.3

0.4

0.5

2195 2200 2205 2210 2215 2220

S w

( f r a c

t i o n

)

Depth (m)

Core Sw Archie Sw Default Archie Sw Optimized

Fig. 17. Comparison of log-derived S w using Archie’s equation using composite m , nparameters derived from Area A core (red curve) and best-t m, n parameters. Coredata (blue triangles) are from Area B. (For interpretation of the references to colour inthis gure legend, the reader is referred to the web version of this article.)



12/25

sample inhomogeneity due to spatial spread of the samples;

samples tend to be biased toward more mechanically competentmaterials; fresh core should be used (versus legacy core);commercial labs may not divulge all the details of the testing(number of strain gauges used, orientation etc.); there are miner-alogic controls on mechanical properties of shales; and conningpressure impacts Poisson’s ratio. Bustin et al. (2008b) noted thatPoisson’s ratio and Young’s modulus are a function of mineralogyand fabric which, in turn, are functions of sedimentology, prove-nance, diagenesis, and tectonics. Barree et al. (2009) further notethat saturation, stress history and anelastic strain (amongst other)factors affect static and dynamic measurements of elastic moduliifor UGRs e they also urge caution when attempting to scale-upcore-derived properties.

Clearly a goal of rock mechanical property measurements is to

ascertain which portions of the reservoir are amenable to hydraulicfracturing (more brittle). Britt and Schoef er (2009) furtheremphasize the importance of mineralogy on shale rock mechanicalproperties and suggest that if shales contain >35e40% clay, they arenot likely to be prospective. Those authors also recommend triaxialcompression tests (for static Young’s Modulus and Poisson’s Ratio)and ultrasonic velocity tests for dynamic measures of those prop-erties, which in turn can be used to distinguish brittle versusductileshale (Rickman et al., 2008).


No rock mechanical property testingwas available foreitherof ourtight gas study areas, but rock mechanical properties inferred fromopenholelogs yield a highYoung’s Modulus,>4.5107 kPa(6.5106

psi), and low Poisson’

s ratio (w

0.2), suggesting that the rock should

be easily fractured (all other factors considered). Frac modeling

utilizing these values was discussed in Clarkson et al. (2011a).

3. Well log analysis

UGR log analysis has progressed dramatically over the past twodecades. Excellent reviews for shale gas reservoirs have recentlyappeared (ex. Sondergeld et al., 2010a; Jacobi et al., 2008;Bartenhagen, 20 09), demonstrating that logs, if properly cali-brated against core data (with its inherent uncertainties), can bea useful technique for estimating: properties to estimate free-gasstorage (porosity and S w for example); total organic carboncontent (for estimating sorbed gas content); rock mechanicalproperties; and natural fracture properties. We refer to the listedreviews for in-depth coverage of these advances, including the

complications and corrections required. One particular compli-cation, which is generally consistent for all unconventionalreservoir types, is the issue of resolution of logs relative to theheterogeneity often exhibited at the core scale. Core measure-ments are often performed on sample sizes signicantly below logscale, so there needs to be some attempt to scale up the coremeasurements in a meaningful way to allow for calibration of logto core data. We discuss core-log calibration using one of our tightgas study areas.

3.1. Tight gas examples

Forour tight gas example, we concentrate on the issue of scale-up for core-log calibration in the nely-laminated reservoir of

Area B and assume that most of the gas is stored in the free-gas

Fig. 18. Comparison of EL and IAS models: equilibrium composition diagrams for the various binary mixtures. IAS predictions for two different total pressures are compared.Modied from Hartman et al. (2011).



13/25

state. Limited TOC data in the area suggest that TOC is generally


14/25

example, wireline bulk density (RHOB) and gamma ray (GR) SVs(Fig. 23) show different characteristics for the 2198e2218 m

interval. The steadily increasing semivariance of the density SV reects the trend of decreasing porosity with depth (Fig. 24). TheGR SV, however, shows holes (decreases in the SV) at lags of approximately 2.6, 5.2, and 7.8 m, suggesting a strongly cyclicelement to the GR responsewith wavelength of 2.6 m. It is commonin turbidite deposits to nd meter-scale cyclicities and may be dueto lobe shifts or variation in sediment input (e.g., Berg, 1986).

Depending on what factors are controlling the formationpermeability,we could see eithera simpletrend orcyclicity or both inthe permeability variations in this interval. Analysis of the probepermeameter measurements reveals a cyclic element (Fig. 12).Fourier transform (FT) amplitudes of the probe data, averaged overa span of 7 cm,showa largepeak at approximately 2.6m wavelength,similar tothat observed intheGR FTbut not the RHOBFT (Fig. 25).AllFTs also show a smaller peak at approximately 1e1.5 m wavelength.Cyclicities at these wavelengths are often dif cult to identify in coreor logs, and might be mistaken for random variations in the SVs. Theprobe FT also shows there is very little cyclicity at wavelengthsbetween 0.2 and 1 m. Cyclicities at wavelengths smaller than15 cm, which might be present from lamination-scale variations,are ltered out by the 13-point (30 cm) running average applied tothe probedata.The smallamplitude of the probeFT amplitudebelow1 m suggests that theprobe data canbe averaged over50 cm or largerlengths without loss of detail when scaling up the permeability forow simulation models or correlations with other measurements.That is, except for variations at the lamination (cm) scale, averaging

the probe data over 50 cm intervals will not result in a loss of

information when scaling up.

4. Pre- and post-frac well test analysis

Well-test analysis can be performed before and after hydraulic-fracture stimulation. Pre-frac tests are generally designed to obtainreservoir information, such as initial pressure and permeability, toaid in stimulation design. Post-frac analysis is generally designed toestimate hydraulic-fracture properties (frac half-length andconductivity) and, if a radialow period is reached, an independentestimate of permeability. Post-frac analysis for tight gas and shalegas reservoirs often does not yield an estimate of permeabilitybecause of the excessive times required to reach a radial owperiod (due to very low permeability). Often operators are not

willing to shut-in wells for extensive periods of time, so recentlythere has been an emphasis on the use of tests where the testingperiod is much shorter (ex. pre-frac tests). However, as we willillustrate with our tight gas example, if estimates of initial pressureand permeability can be derived from independent sources, thenpost-frac tests can be quite useful for obtaining quantitative esti-mates of hydraulic fracture properties, provided efforts are taken toreduce wellbore storage. Pre-frac testing (openhole and cased hole)can have a relatively short test time, and therefore a strongerchance of yielding permeability information, but generally sufferfrom a small radius of investigation compared to longer post-fractests or rate-transient analysis (discussed later).

Of course there are many reservoir-related issues associatedwith UGR that complicate interpretations of well-tests, which will

be discussed in more detail during the rate-transient section below.Specialized procedures for dealing with multi-phase ow anddesorption, multi-layer behavior and dual porosity behavior arediscussed elsewhere in the context of coalbed methane reservoirs(Clarkson and Bustin, 2011; Mavor and Saulsberry, 1996). The focusof this short discussion will be on pre- and post-frac testing of tightand shale gas reservoirs.

4.1. Pre-frac testing: diagnostic fracture injection tests (DFITs)

Over the last ve decades, DFITs (also called mini-fracs or pre-frac diagnostics) have been performed prior to stimulation treat-ments to diagnose many important fracturing characteristics priorto a nal frac design (Cleary et al., 1993; Nolte and Smith, 1981;

Barree, 1998; Nolte, 1990). DFITs involve injecting

uid volumes

Fig. 21. Winland-style plot of corrected (for in-situ stress) probe permeability dataversus density porosity along with the lines based on several values of r p35 for Area B.The range of pore widths from N2 adsorption analysis is also given. Modied fromClarkson et al. (2012).

Fig. 22. Winland-style plot of routine permeability data versus helium porosity along

with the lines based on several values of r p35 for Area A.

Fig. 23. Semivariograms of wireline GR (left axis) and density (right axis) from interval2198e2218 m in tight gas Area B.



15/25

(small relative to the main stimulation) into the reservoir (typicallyabove fracture gradient) to characterize hydraulic fracture-relatedreservoir properties. In more recent times, evaluation of a DFIThas evolved to include an estimate of permeability via the After-Closure Analysis (ACA) approach (Abousleiman et al., 1994).

ACA relies on the assumption that the uid injected into thereservoir will ultimately remain stationary in the reservoir suchthat, in late time, the only uid moving in the reservoir is thereservoir uid (e.g. gas). ACA involves the identication of pseudo-

radial ow to determine reservoir transmissibility (kh/ m). Accurateestimation of k and kh is of importance in the fracture designprocess since they can strongly inuence the economicallyoptimum fracture conductivity and fracture half-length. In someenvironments the results of ACA may determine if a zone should bestimulated at all. Estimates of permeability are critical to assess theeffectiveness of the stimulation, post-treatment. Numerous papershave tested and validated this technology in the eld (Ramurthy

et al., 2002; Chippereld and Britt, 2000; Nolte, 1997; Gulrajaniet al., 1998; Talley et al., 1999; Benelkaldi, 2001). Strengths andweaknesses of the DFIT technique are listed below:

Strengths

DFITs have an advantage over more conventional testingmethods in terms of cost. Conventional methods (DST andpressure buildups) often require additional cost both in termsof equipment requirements and delays in well ‘on-line’ dates.

Conventional gas well testing is also very limited in lowpermeability reservoirs, since these wells often produce at verylow rates pre-frac. In contrast, if a DFIT is already beingundertaken as part of the completion operations, the incre-mental cost of performing ACA is often less than conventionalwell-test methods since there is no requirement for additionalequipment and well ‘on-line’ dates are not further postponed.In addition ACA, since it is an injection test evaluation, does notrequire the well to ow pre-frac for analysis.

DFITs are multi-purpose; as with all short-term transient tests,DFITs can be used to determine other sources of reservoirinformation such as the presence of natural fractures(Chippereld, 2005). In addition to permeability, these testscan also be used to determine reservoir pressure from the

pseudo-linear ow regime (Talley et al., 1999).

Weaknesses

The Requirement of a Small Fluid Bank: Unless this is the case(i.e. a small injection), the after-closure ow regimes may bemasked or delayed due to a large uid bank. If this is the case,and uid remains mobile in the reservoir, the permeability canbe underestimated (in the case of gas) if one mis-interpretsradial ow from the mobile water as representing the trans-missibility of the native reservoir uid.

Limited Injection Interval: This technique, as with conventionalwell testing, assumes the pressure disturbance occurs overa known interval. To maximize the chance of this occurringDFITs are best conducted on single sands and above fracture

2500

2520

2540

2560

2580

2600

2620

2640

2660

2680

27000

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

2198 2200 2202 2204 2206 2208 2210 2212 2214 2216 2218

B u

l k D e n s

i t y

( k g / m

3 )

P o r o s

i t y

( F r a c

t i o n

)

Depth (m)

Core Porosi ty Log Bulk Densi ty

Fig. 24. Wireline bulk density and full-diameter core (routine) porosity versus depth in tight gas Area B. Modied from Clarkson et al. (2012).

Fig. 25. Fourier transform amplitudes of probe permeameter (left axis) and wireline

GR and density (right axis) from interval 2198e

2218 m in tight gas Area B.



16/25

gradient (to ensure connection to the entire sand). This tech-nique obviously lends itself to targeted stimulation approaches(Beatty et al., 2007).

Shut-In Period: The shut-in period required to describe all theow regimes in ACA can be lengthy. Several papers have alsoshown however that this analysis approach can achieve goodapproximations of formation parameters without completelyobserving all the ow regimes (Chippereld, 2005).


Although we may seem to have been critical of the use of conventional post-fracture well-test analysis (such as ow andbuildups) in tight gas and shale gas reservoirs, we believe thesetests can still be very useful for obtaining hydraulic fracture infor-mation, provided efforts have been made to properly cleanup thewell after fracturing operations, and efforts have been made toreduce wellbore storage. Typically, only bilinear or linear owperiods are present in the data for tight gas wells, which can beused to obtain an estimate of fracture conductivity and fracturehalf-length, respectively, if permeability is known. Althoughpseudo-radial ow is unlikely to be reached during the test in many

tight gas cases, if an independent estimate of reservoir pressure isavailable (e.g. a pre-frac test), as well as permeability (e.g. rate-transient analysis, core or pre-frac test), then the post-frac test isirreplaceable as a way to quantify hydraulic fracture properties. Anexample is given in Fig. 26 for a hydraulically-fractured vertical wellin Area B. The well was shut-in for two weeks and exhibitsa hydraulic fracture signature (linear ow), but no radial ow eassuming that radial ow is reached at the end of the ow periodcauses permeability to be overestimated (0.084 md), and hydraulicfracture half-length to be underestimated (91 ft). The permeability

estimate from rate-transient analysis for this same well (afterproducing for 600 days), is 0.016 md, which is still considered to beoverestimated because radial ow is not thought to have beenreached during the production period either. However, using theow capacity kh (and with an assumption of h and k) derived fromrate-transient analysis, a new estimate of hydraulic-fracture half-length may be derived (w200 ft), which is much closer to the RTA-derived values of 250e300 ft (discussed below). The somewhatshorter well-test derived values may be due to incomplete cleanupof the fracture post-stimulation. Nonetheless, we think that theshort, post-frac well-test is justied, provided constraints on initialpressure and reservoir permeability are available, to provide anindependent estimate of hydraulic-fracture properties. Periodictesting of this nature could allow for effective half-length andconductivity to be monitored, perhaps for the purposes of identi-fying re-stimulation candidates. It should be noted that in our tightgas examples, a low-complexity fracture (planar) fracture systemembedded in a single-porosity reservoir was assumed in theanalysis, which is consistent with microseismic observations (seeClarkson and Beierle, 2011). However, in shale gas reservoirs,a natural fracture network may have to be considered in the anal-ysis (Ozkan et al., 2009; Apaydin et al., 2011).

Little pre-frac well-test was available in either tight gas studyarea (Area B or A), and what was available was of variable quality,mainly due to inadequate test durations that did not allow pseudo-radial ow to develop. Some exceptions exist, however, such as theDFIT test interpretation shown in Fig. 27, which exhibits anapparent radial ow period that could be interpreted for a kh andpressure value. The kh estimate for this well (0.182 md*m or 0.60md*ft) is within the range of values obtained from rate-transientanalysis (Clarkson and Beierle, 2011) and appears to be reason-able. Also, the estimated bottomhole pressure is consistent with

Fig. 26. Example post-frac buildup analysis of a hydraulically-fractured vertical well in tight gas Area B. Note the short buildup times. A hydraulic fracture signature is evident froma formation linear ow period. The estimate of formation permeability is too high in this case (radial ow not reached), and hence hydraulic fracture half-length is underestimated.

A more robust estimate for fracture half-length can be obtained using the RTA-derived permeability values.



17/25

estimates from conventional well-tests. As discussed above, idealpre-frac tests can be used to obtain an independent estimate of initial pressure and permeability that can be used for analysis of short post-frac ow and buildup tests.

ACA can be performed via several different approaches.Chippereld (2005) used the NDP (Nolte Diagnostic Plot)(Chippereld, 2005) to determine permeability-thickness of a tightgas reservoir and to identify natural fractures, which appeared asa dual porosity dip on the NDP plot. The interpretation was alsosupported by pre-frac pressure buildup and core analysis(Chippereld, 2005). Recently, Mohamed et al. (2011) proposed an

analysis approach for ACA that utilizes more conventional pres-sure-transient analysis methods.

Although the intent of the above examples is to demonstratehow DFITs can be used to estimate permeability and initialreservoir pressure, we note that DFITs can also be useful indetermining the appropriate fracture treatment in unconven-tional gas reservoirs. Ramurthy et al. (2011) recommended a seriesof testsfor shale gasreservoirsto help determine whether fracturestimulation treatments should be designed to maximize surfacearea (“Barnett-type waterfracs”) or conductivity. They note thathigh-volume waterfracs may not be the optimal stimulationchoice for some shale plays, particularly when complex(“dendritic”) fractures cannot be created, which in turn dependson in-situ stress anisotropy, fabric of the rock etc. Leakoff behavior,

time to reach closure, closure pressure, permeablility and otherproperties that can be determined from DFIT tests can be used tohelp identify reservoir type, and its suitability for different frac-ture treatments.

5. Rate-transient and production-data analysis

Rate-transient and production data analysis have undergonea rapid evolution recently in an attempt to make these methodsapplicable to UGRs. Both analytical and empirical methods havebeen advanced. In previous work (Clarkson et al., 2009; Clarksonand Beierle, 2011), we categorized production analysis methodsinto: a) straight-line (ow-regime) methods; b) type-curvemethods; c) analytical and numerical simulation and d) empirical

methods. Straight-line analysis involves

rst the identi

cation of

ow regimes, followed by analysis of the ow regime data onspecialty plots to obtain hydraulic fracture or reservoir propertyinformation. Type-curve methods involve matching productiondata to analytical and/or empirical solutions to ow equationscast in dimensionless form. Analytical and numerical simulatorscan be “calibrated” to dynamic data (rates and pressures) to deriveproperties. Empirical equations can be t to production datato yield a forecast. Lastly, there is a new class of forecasting tech-niques referred to “hybrid” methods which combine analyticalsimulation for transientow and empirical methods for forecastingboundary-dominated ow. For shale gas reservoirs, these hybrid

techniques have been developed for the simple case of a homoge-nous completion (Nobakht et al., 2010), where all hydraulic frac-tures emanating from a horizontal well are all the same length, andfor heterogeneous completions (Nobakht et al., 2011a), where thehydraulic fractures may be of varying length.

When performing quantitative production analysis of UGRs, theanalyst can encounter a wide range of reservoir characteristics thatmay need to be accounted for (or at least acknowledged) in theanalysis including:

Low matrix permeability, which causes transient ow periodsto be extensive

Dual porosity or dual permeability behavior, due to existence of natural fractures or induced hydraulic fractures (or both)

Other reservoir heterogeneities, such as multi-layers (inter-bedded sand/silt/shale) and lateral heterogeneity Stress-dependent permeability, due to a highly-compressible

fracture pore volume Desorption of gas from the organic matrix Multi-mechanistic (non-Darcy) ow in shale matrix, caused by

gas-slippage along pore wall boundaries and diffusion (see Javadpour, 2009),orinthehydraulicfracturesduetoinertial ow

Recent advances in these methods specically for UGRs include:

1. Type-Curves developed for hydraulically-fractured wells Innite- and nite-conductivity fractures (ex. Pratikno et al.,

2003; Agarwal et al., 1999)

Elliptical

ow (Amini et al., 2007)

Fig. 27. Example after-closure analysis (ACA). In this example, the radial ow period is analyzed to obtain a permeability estimate.



18/25

Shale gas reservoirs (Lewis and Hughes, 2008; Moghadamet al., 2010; Nobakht et al., 2011b; Abdulal et al., 2011)

2. Straight-line (ow-regime) techniques adapted To analyze ow regimes in unconventional gas reservoirs

(ex. Wattenbarger et al., 1998; Clarkson et al., 2009; Belloand Wattenbarger, 2008)

3. Type-curve and straight-line methods modied to account for: Desorption (Clarkson et al., 2007; and Gerami et al., 2007) Multi-phase ow (Mohaghegh and Ertekin, 1991; Clarkson

et al., 2009) Non-static permeability (Thompson et al., 2010) Non-Darcy ow (Clarkson et al., 2011b)

4. Numerical and analytical simulators for shale5. Improvements in ow-regime identication (ex. Ilk et al.,

2005)6. New empirical methods (ex. Power-Law Exponential) (Ilk et al.,

2008)

Strengths and weaknesses of the RTA/PDA technique are listedbelow:

Strengths

Wells do not need to be shut-in; RTA/PDA does not requireplanned shut-ins. As a result there is no deferral of revenue tocomplete the analysis.

Data commonly available; Gas production data are alwayscollected for producing wells, whereas shut-in tests are notcommonly available. Although the technique can be used withwellhead data alone (with an estimate of sandface pressurefrom a wellbore model), the information extracted can beimproved using downhole gauge data, which are notcommonly available. Downhole gauge data should be consid-ered by operators to improve the quality of the analysis.

Amenable to Gas; RTA/PDA allows the user to account for gasproperties unlike Arps and other traditional declineapproaches that assume constant incompressible uid with

a xed bottom-hole pressure. Amenable to Tight Formations; In many tight gas reservoirs the

time required to observe allow regimes can be very long. This

makes RTA/PDA a more practical testing method than othertraditional approaches (ex. pressure buildup tests (PBUs)).

Non-Idealities can be observed in the data; While RTA/PDA isbased on simple single-layer simulation models, many ‘non-idealities’ can be observed and diagnosed in the data(Fig. 28).

Weaknesses

Data can be of poor quality with range of outcomes possible ; Insome cases the results from RTA/PDA can have a high degree of uncertainty. The example provided below (Fig. 29) showsnumerous outcomes that can match the data. One family of results would suggest that with modied stimulation design,improvements could be achieved by increasing the effectivefracture length (Family 1). The other family of matches (Family2) would suggest that the eld production cannot be improvedwith existing completion methods. Understanding that thisuncertainty exists is key as it will decide whether additionalinformation (ex. PBU) is necessary before business decisionscan be made.

Boundary dominated ow may never exist ; This can create

problems with reserve estimation because quantitative esti-mates can only be derived if boundary-dominated ow isreached. Tight formations may exhibit transient ow for longperiods of their life. A conservative approach to reserve esti-mation would be to assume that, while the well is still intransient ow, that boundary-dominated ow would occurshortly after the existing production period, and imposea drainage area corresponding to the contacted drainage areareached up until that point. If this is not done, then extrapo-lation of transient ow data can yield EUR over-estimates(Rushing et al., 2007). In layered reservoirs, some higherpermeability zones may reach boundary-dominated owrelatively quickly; in these scenarios, reserves may be under-estimated if long-term transient ow of lower-permeability

layers is not taken into account. Finally, false boundary-dominated ow signatures may be seen for multi-fracturedwells completed in tight formations due to fracture

Fig. 28. Production data analysis examples showing how non-idealities appear on type curves of log(q/dP) versus material balance pseudo-time.



19/25

interference or boundaries of the stimulated reservoir volumebeing reached, potentially leading to reserves underestimation.

Recently, rate-transient analysis methods have been used to

analyze shale gas reservoirs to extract information about contactedsurface area generated from hydraulic fracturing operations, andpossibly hydraulic fracture spacing (Anderson et al., 2010). We notethat RTA used on its ownwith no further information about fracturegeometry could be very misleading. Whenever possible, productionanalysis should be combined with other surveillance data such asmicroseismic, production and tracer logs, so that ow regimesencountered during the analysis can be interpreted for physicalmeaning. In the tight gas example that follows, we discuss inte-gration of data for a more meaningful analysis e for shale gasreservoir examples, we refer to Anderson et al. (2010).

5.1. Tight gas examples

In a previous paper (Clarkson and Beierle, 2011), hydraulic-fractured vertical and horizontal wells were analyzed using rate-transient techniques, including straight-line, type-curve andanalytical and numerical simulation in Area B. A detailed procedurefor analysis was provided, emphasizing the integration of surveil-lance data such as microseismic, production and tracer logs. Rate-transient analysis of vertical wells was rst performed to estab-lish estimates of reservoir (ex. kh) and hydraulic fracture (effectivelength and conductivity) properties, prior to analysis of multi-fractured horizontal wells, which were used in later develop-ment. The consistency in stimulation treatments for both verticaland horizontal well hydraulic fracturing stages allowed for directcomparison of results. Both commingled and individual zoneproduction rates were analyzed in multi-fractured horizontal wells.

Detailed core and log analysis for Area B, discussed in previoussections, provided volumetric data as input, and very importantly,estimates of matrix permeability for comparison with RTA. Further,some pre-frac well-test analysis was performed on vertical wells inthe area to provide for additional comparisons.

Straight-line analysis examples for a vertical and horizontal well(commingled stage) production data are summarized in Fig. 30aand b, and type-curve analysis in c and d. Flow-regime identica-tion was rst performed using semi-log and linear derivatives andother diagnostic plots (not shown) to ensure the correct owregimes were being analyzed. We note that plots containingsuperposition time functions contain a bias towards the owregime assumed for superposition calculations, and should not beused for ow-regime identication. A planar hydraulic fracture

model was chosen for the type-curve model due to the

low-complexity of hydraulic fracture geometry as inferred frommicroseismic data (Clarkson and Beierle, 2011). Only linear owanalysis is shown, although (early) bilinear analysis was performedto estimated fracture conductivity for type-curve selection, andowing material balance to obtain a minimum contacted gas-in-place (wells are still in transient ow). Additionally, althoughradial ow did not appear to be fully-developed in vertical wells,a maximum kh value was derived, and resulting permeabilities(with estimation of h) used in horizontal well analysis. The effectivefracture half-length and conductivity from vertical well analysis issimilar to that estimated from horizontal wells (noting that a totalhalf-length from all stages is derived from the horizontal wellanalysis, and per stage values were derived from number of stages).This consistency makes sense due to the similarity in fracture stagetreatments for verticals and horizontal wells. The half-lengths fromlinear ow analysis are assumed to be overestimated, due to use of pseudotime values in linear superposition time calculations thatare anchored to pore volume average pressure; as noted by

Nobakht and Clarkson (2011a,b), a corrected pseudotime yieldsmore accurate values.

Type-curve estimates of hydraulic fracture lengths were slightlysmaller for vertical wells e this is because the well is not yet inboundary-dominated ow, and hence ultimate drainage radius isunderestimated. The use of material balance time combined withliquid loading behavior is believed to cause a false-boundarysignature. The selected type-curve stem is the ratio of drainageradius to hydraulic fracture half-length, so if drainage radius isunderestimated, so is hydraulic fracture half-le

Documents

Unconventional Gas Reservoir Evaluation. What Do We Have to Consider - Clarkson, 2012