SPE Copyright 2007, Society of Petroleum Engineers

Transcription

1 SPE Beyond Decline Curves: Life-Cycle Reserves Appraisal Using an Integrated Work- Flow Process for Tight Gas Sands J.A. Rushing, SPE, Anadarko Petroleum Corp., K.E. Newsham, SPE, Apache Corp., A.D. Perego, SPE, Anadarko Petroleum Corp., J.T. Comisky, SPE, Apache Corp., and T.A. Blasingame, SPE, Texas A&M University Copyright 2007, Society of Petroleum Engineers This paper was prepared for presentation at the 2007 SPE Annual Technical Conference and Exhibition held in Anaheim, California, U.S.A., November This paper was selected for presentation by an SPE Program Committee following review of information contained in an abstract submitted by the author(s). Contents of the paper, as presented, have not been reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material, as presented, does not necessarily reflect any position of the Society of Petroleum Engineers, its officers, or members. Papers presented at SPE meetings are subject to publication review by Editorial Committees of the Society of Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of the Society of Petroleum Engineers is prohibited. Permission to reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O. Box , Richardson, Texas U.S.A., fax Abstract Decline curve analysis is often either the only or the primary tool used for reserve evaluations in tight gas sands. However, the flow and storage properties characteristic of lowpermeability sands often preclude accurate assessments using only or primarily decline curve analysis, especially early in the productive life. The most accurate reserve estimates incorporate multiple data sources and the appropriate evaluation techniques. Therefore, this paper presents a reserves appraisal work-flow process that complements traditional decline curve analyses with comprehensive and systematic data acquisition and evaluation programs that integrate both static and dynamic data. Our approach which has been developed specifically to incorporate the production characteristics of tight gas sands is an adaptive process that allows continuous but reasonable reserve adjustments over the entire field development and production life cycle. Implementing this process will prevent unrealistic (either too low or high) reserve bookings. Although it is applicable during any field development phase, our work-flow process is most beneficial during early stages before true boundary-dominated flow conditions have been reached and when reserve evaluation errors are most likely. Introduction Estimating reserves in any type of hydrocarbon resource is an important process for all publicly traded oil and gas companies. The reserve estimating process is necessary for assessing the value of both individual producing properties as well as the entire company. Accurate reserve assessments are also extremely critical since the reserves value may have a direct impact on company earnings and the overall balance sheet. Moreover, the value of booked reserves may influence the cost and availability of capital required for a company s future growth. Perhaps most importantly, reserves assessment is a legal requirement to satisfy reporting and disclosure requirements set forth by various regulatory and governmental agencies. The most widespread reserve estimating techniques historically used by the oil and gas industry were initially developed for conventional oil and gas reservoirs 1-7 but have been routinely extended (sometimes inappropriately) to unconventional reservoirs. The most common reserve estimating methods include: curve fitting existing production data using the traditional Arps 8 decline curve techniques and extrapolating the model to estimate remaining reserves and production at some future economic conditions; calculating original volumetric gas-in-place and applying a recovery factor to estimate reserves. Recovery factors are estimated from field analogues or computed using a model (i.e., reservoir simulation, material balance, etc.); using conventional material balance models to estimate original gas-in-place and applying a recovery factor to estimate reserves and future production; and history matching well and/or field production with a reservoir simulator, and estimating future production and reserves with the calibrated model. Although the oil and gas industry has long recognized that the most accurate reserve assessments are made by integrating multiple analysis techniques, the most widely used method especially for fields with sufficient production histories has typically still been just decline curve analysis. Depending on the type, quantity and quality of available data, decline curve analysis is sometimes the only method available for evaluating reserves. We should note that, when applied under the correct conditions, decline curves are excellent tools for estimating reserves in conventional oil and gas resources. Unfortunately, many of us in the industry either have forgotten or have chosen to ignore the conditions under which use of the Arps 8 decline curves are appropriate. Application of a decline curve methodology using the Arps models implicitly assumes the following: 9 extrapolation of the curve fit through the current or historical production data is an accurate model for future production trends;

2 2 J.A. Rushing, K.E. Newsham, A.D. Perego, J.T.Comisky, and T.A. Blasingame SPE there will be no significant changes in current operating conditions or field development that might affect the curve fit and the subsequent extrapolation into the future; the well is producing against a constant bottomhole flowing pressure; and well production is from an unchanging drainage area with no-flow boundaries, i.e., the well is in boundary-dominated (stabilized) flow. Violation of one or more of these assumptions may cause significant errors in reserve estimates. In fact, Thompson, et al. 10 and Wright 11 have documented that, even in conventional reservoirs, there are often significant uncertainties and errors in reserves estimates derived from traditional decline curve evaluations. Wright 11 has further demonstrated the percentage error in remaining reserves is larger and more uncertain than the percentage error in ultimate recovery, and that inclusion of more production data does not necessarily either improve the accuracy or reduce the uncertainty. The errors in reserves estimates derived from a decline curve methodology are often exacerbated in tight gas sands. Similar to conventional oil and gas systems, tight gas sands are characterized by complex geological and petrophysical systems as well as heterogeneities at all scales. However, unlike conventional reservoirs, tight gas sands typically exhibit unique gas storage and producing characteristics that make it difficult to evaluate reserves accurately with conventional techniques, especially before true boundarydominated flow conditions have been achieved. For example, many wells producing from low-permeability gas sands are completed in multiple layers, each having different flow and storage capacities. Individual layers may also have significantly different initial reservoir pressures, particularly if separated by large vertical intervals. Moreover, to reduce completion costs, production from each layer is frequently commingled in the same wellbore. Depending on the permeability contrast, each layer may reach true boundarydominated flow at different times during the well s producing life, thus making it impossible to use an Arps decline curve methodology to estimate reserves, especially if we cannot accurately allocate gas production to each layer. Because of the reservoir complexities associated with lowpermeability reservoirs and combined with the inherent uncertainty of traditional decline curve analysis, we have concluded that accurate reserves assessment in tight gas sands cannot be achieved only or primarily with decline curve analysis but must integrate multiple data sources and evaluation methodologies. As a result, we have developed a reserves appraisal work-flow process that systematically integrates multiple static and dynamic data sources and types (i.e., engineering, geological, and petrophysical data) at varying scales and provides appropriate evaluation techniques for different data types. Our approach is an adaptive process that allows continuous but reasonable reserves adjustments over the entire field development life cycle. Unlike probabilistic reserve evaluation techniques that attempt to quantify uncertainty, our method is completely deterministic. The hypothesis underlying our work-flow process is that we can reduce both the inherent uncertainties and errors in reserves assessment by complementing (rather than replacing) traditional decline curve analysis with application of consistent, systematic, and comprehensive data acquisition and appropriate evaluation programs. Overview of Reserves Appraisal Work-Flow Process Our reserves appraisal work-flow process (Fig. 1) has two fundamental stages quantification and validation. Reserves from producing wells are initially quantified using a traditional decline curve methodology. Depending on field maturity, data quality and quantity, the Arps 8 decline curve parameters (i.e., initial decline rate, D i, decline exponent, b, etc.) may be determined either from existing producers in the field or estimated from suitable field analogies. Static Pressures Core Acquisition & Evaluation Well Log Acquisition & Evaluation Fluid Acquisition & Evaluation Quantification Validation Validation Volumetric GIP (Static) Traditional Decline Curve Analysis Reserves Validation Validation Contacted GIP (Dynamic) Well Surveillance & Monitoring Well Performance Analysis Reservoir Simulation Pilot Infill Drilling Figure 1 Schematic diagram showing various stages of workflow process for evaluating reserves in tight gas sands The next step in our work-flow process is a reserves validation stage in which reserve volumes are substantiated by comparison to both volumetric and contacted gas-in-place volumes. We define volumetric gas-in-place (VGIP) as a static volume representing the total resource-in-place, while the contacted gas-in-place (CGIP) is a dynamic volume representing the fraction of VGIP in pressure communication with producing wells at any point in time. Note that the CGIP will increase with time as the drainage areas of individual wells increase during both transient flow and the transitional period between the end of transient and the onset of boundary-dominated flow. Eventually, however, the CGIP will cease to change as the flow conditions reach true boundary-dominated flow. The onset of boundary-dominated flow may be caused either by no-flow reservoir boundaries (i.e., sand discontinuities caused by faults, pinch outs, flow barriers etc.), by well interference from adjacent producers, or a combination of both conditions. Accurate quantification of VGIP requires integrated description programs of static reservoir data, including reservoir pressure histories, core and fluid studies, and well logging programs. The overall objective of these programs is to describe the reservoir container in sufficient detail so that we might estimate the total resource in place. Similarly, the CGIP is characterized by integrating dynamic reservoir data (e.g., flowing well pressures, gas and water production

3 SPE Beyond Decline Curves: Life-Cycle Reserves Appraisal Using an Integrated Work-Flow Process for Tight Gas Sands 3 histories, etc.) with multiple evaluation techniques, including production performance analyses, reservoir simulation, and/or well surveillance programs. Since the CGIP volume is dynamic (i.e., changes with time, production, and field development maturity), the overall objective of this component is to determine how much of the VGIP is in pressure communication or contact with the producing wells at any given time during the field development and production history. Realistic and consistent reserve estimates may be achieved by limiting or constraining reserve volumes to a fraction of both VGIP and CGIP. Further, the constraints are dictated by the field development stage and/or the flow conditions (i.e., transient, transitional, or boundary-dominated). Generally, we would expect to see specific relationships among various gas volumes, including cumulative gas production (G p ), estimated ultimate recoveries (reserves or EUR), contacted gas-in-place (CGIP) and volumetric gas-in-place (VGIP) at various stages of the field development. Knowledge of these relationships as well as the type of flow period can help guide us in our reserve additions and prevent unrealistic (either too low or high) reserve additions. The remainder of this paper describes in detail each of the stages in our reserves appraisal work-flow process. We also provide procedures on how to employ the work-flow, including guidelines on constraining reserve estimates by comparing all of the various gas volumes. Reserves Quantification Stage: Decline Curve Analysis The first stage in our work-flow process illustrated in Fig. 1 is the quantification phase during which we use traditional decline curve analysis techniques to estimate reserves. Decline curve analysis consists of plotting production rate against time, history matching the production data using one of several industry-standard models i.e., the Arps 8 exponential, hyperbolic or harmonic decline models and extrapolating the established trend into the future. Decline curve analysis as historically practiced by the industry is quite simply a curve fitting process which does not have a theoretical basis. An exception to this statement is exponential decline which can be derived theoretically from a single-well model producing at a constant bottomhole flowing pressure during boundary-dominated flow conditions. Traditional Arps Decline Curve Models. Early attempts at decline curve analysis focused on finding plotting techniques or functions (i.e., models) that linearized the production history since linear functions are very easy to manipulate mathematically and/or graphically. 9 Accordingly, the most common traditional decline curve analysis technique is a plot of the logarithm of production against time, i.e. a semilog plot which is sometimes called exponential or constant-percentage decline. Subsequent work by Arps, 8 however, demonstrated that the production performance of all wells cannot be modeled with simple exponential decline, so he proposed both harmonic and hyperbolic decline functions. All traditional decline curve analysis is based on the general form of the Arps 8 empirical rate-time decline equation, q( t ) qi = (1) 1/ b ( 1+ bd t) i where D i is the initial decline rate, q i is the gas flow rate, and b is the Arps decline curve constant or exponent. Equation (1) has three different forms exponential, harmonic, and hyperbolic depending on the value of the decline exponent, b. Each of these equations has a different shape on Cartesian and semilog graphs of production rate vs. time and cumulative production. Because of these differences, the various curve shapes can help identify the type of decline for a well if boundary-dominated flow conditions have been reached. Exponential or constant-percentage decline which is a special case of Equation (1) when the b-exponent equals zero is characterized by a decrease in production rate per unit of time that is directly proportional to the production rate. The exponential decline equation can be derived from Equation (1) with a b-exponent of zero as qi D t q t) = = q i D t ie e i (... (2) Similarly, harmonic decline is the special case of Equation (1) when the b-exponent equals one, and can be written as q( t) = qi... (3) ( 1+ D i t) Another decline curve model suggested by Arps is hyperbolic decline which can be derived from Equation (1) when b is between 0 and 1.0, or q( t) = qi... (4) ( 1+ bd ) b i t 1/ We should note that the value of b determines the degree of curvature of the semilog decline, ranging from a straight line with b=0 to increasing curvature as b increases. Although he did not base his conclusions on theoretical considerations, Arps 8 suggested the b-exponent should fall between zero and one but offered no discussion of the possibility that b might be greater than one. Moreover, Arps did not attempt to correlate reservoir or fluid properties to the value of b. Fektovich, et al. 12 summarized relationships between the b- exponent and various types of reservoir heterogeneities and drive mechanisms, but they did not suggest b could ever exceed one. Gentry and McCray, 13 however, simulated a number of cases for an oil reservoir and concluded that b could indeed exceed one, particularly for reservoirs with layers having different properties. Maley 14 also documented cases from the Lower Cotton Valley tight gas sands in Louisiana with b-exponents greater than one. More recently, Rushing, et al. 15 conducted a number of simulation studies for tight gas sands at high-pressure/hightemperature (HP/HT) reservoir conditions, and concluded that b would generally fall between 0.5 and 1 for a range of reservoir and hydraulic fracture heterogeneities. They also

4 4 J.A. Rushing, K.E. Newsham, A.D. Perego, J.T.Comisky, and T.A. Blasingame SPE observed that values of b greater than one reflected transient or transitional rather than true boundary-dominated flow. Estimating Decline Curve Parameters. Accurate estimates of remaining reserves and production may be obtained from an extrapolation of the Arps decline curve models if we have correct estimates of decline curve parameters, primarily the b- exponent. There are a number of methods for estimating b, including curve fitting, use of analogs, and application of various types of reservoir models. The applicability of each method depends on the data quality and/or quantity available for estimating reserves. Curve Fitting or History Matching. Curve fitting the historical production with one of the Arps models is the most straightforward and widespread method for estimating decline curve parameters. Accurate evaluation, however, still requires that the well be in boundary-dominated flow; otherwise the curve-fit parameters and subsequent reserve estimates from extrapolation of transient flow or the transitional period between the end of transient and onset of boundary-dominated flow data are certainly not unique and will probably be incorrect. Analogous Fields. We may also estimate decline curve parameters from an analog field, especially if the analog wells are exhibiting boundary-dominated flow. But, we must also be judicious in our selection of an analog field. As discussed by Harrell, et al., 16 and Hodgin, et al., 17 being areally proximal is not necessarily sufficient to justify selection of a field as an analog. Ideally, we must also have similar geotechnical, petrophysical, engineering, and operational parameters to ensure proper and accurate analogies. Table 1 (adapted from References 16 and 17) summarizes some general parameters that should be considered when selecting an appropriate analogous tight gas sand field for estimating reserves. Note that Table 1 is by no means complete, but simply summarizes major parameters that should be considered when selecting an analog field for modeling production performance. Reservoir Simulation. Although the primary objective is typically not to estimate reserves, 18,19 reservoir simulation does offer an alternative for estimating decline curve parameters, particularly when no appropriate analogs are available. A reservoir simulator is essentially a reservoir model that incorporates more detailed and complex engineering and geotechnical data than simpler analytical or semi-analytical models. The typical simulation process is to history match existing production by varying reservoir and/or operating parameters, and then to extrapolate the simulated production profile to some abandonment condition. Both References 18 and 19 suggest that, if a good history match can be obtained, the history-matched model can be considered to be an analog for estimating future performance. But, they also caution against using reservoir simulation alone since the model adequacy depends on the quality and quantity of the data available for history matching. As we have recommended, other types of data and evaluation techniques must be incorporated to validate the simulated production profiles and decline behavior. Production Data Analysis with Decline Type Curves. Decline curve parameters (primarily the b-exponent) can also be estimated using production decline type curve analysis. Reference 20 proposes using the Fektovich type curves 12 and the relationships developed by Chen 21 to first estimate the b- exponent representative of boundary-dominated flow conditions. Holding the b-exponent constant, the other decline parameters are estimated by history matching the production using the Arps models. Table 1 Summary of geotechnical, petrophysical, engineering and operational parameters to consider when selecting field analogs for tight gas sands (adapted from Refs. 16 and 17) Geotechnical Geological age Structural history Depositional environment & stratigraphy Lithology, sedimentology Major heterogeneities, including natural fractures, faults Reservoir continuity ( blanket vs. lenticular ) Reservoir Engineering Reservoir pressure and temperature conditions Reservoir depths Gas, water fluid properties Permeability anisotropy Petrophysical Gross & net sand thicknesses Range of absolute permeability & effective porosity Range of hydraulic rock types Types and extent of rock diagenesis Capillary pressures & vertical fluid saturation distribution Stress-dependent rock properties Operational Engineering Well productivity and well stimulation methods. Well spacing Well architecture (vertical vs. horizontal) Producing constraints and/or enhancements (pipeline capacity, compression, etc.) Data Requirements for Decline Curve Analysis. The accuracy of reserve estimates from traditional decline curve analysis is affected not only by the quality of the recorded production, but also the frequency of data collection. Ideally, we would like to have daily production volumes recorded for individual wells rather than the entire field. We can then evaluate the decline performance using either the recorded daily rates or the computed monthly volumes. In many cases (especially in older fields), we may not have daily production data from individual wells. Rather, we will only have monthly production histories from multi-well leases, production units, or perhaps even an entire field. Allocation of field production to individual wells using well tests is an accepted industry practice, but it is often inaccurate because of infrequent and/or incorrect well test information. Harrell, et al. 16 have also illustrated the errors associated with decline curve analysis using composite field data rather than individual well production. Errors may be compounded if the overall field decline is assumed to be similar to the decline behavior of an average well. Another error source in using a composite field decline is failure to account for future field development and well count, particularly if infill drilling and/or field extension by drilling occurs. Both of these development activities may affect the composite field decline characteristics.

5 SPE Beyond Decline Curves: Life-Cycle Reserves Appraisal Using an Integrated Work-Flow Process for Tight Gas Sands 5 Common Misuses of Decline Curves. Correct application of a decline curve methodology using the Arps 8 models must be made under certain conditions and assumptions that we outlined previously. Unfortunately, these conditions are frequently violated resulting in incorrect reserves estimates, especially in tight gas sands. The most common misapplications of traditional decline curve analysis in lowpermeability gas reservoirs are: Application of the Arps Decline Curves During Transient and/or Transitional Flow Periods: Depending on stimulation effectiveness, effective reservoir permeability and well spacing, the onset of boundary-dominated flow may be delayed for many months or even years in some tight gas sands. The traditional Arps equations are strictly valid when applied to boundary-dominated flow data. There is simply no basis for predicting long-term performance from an extrapolation of the curve fit through data during transient or transitional flow periods. Until all drainage area boundaries have influenced the well s decline characteristics, predictions of long-term behavior are probably both non-unique and incorrect. Failure to Modify Reserve Estimates When Operational Conditions Change: The Arps models assume implicitly that the well is produced at a constant bottomhole flowing pressure condition, 9 so any operational changes that alter this flowing condition could affect the extrapolated decline. For example, many gas fields are placed on compression at various stages of development. Subsequent staged compression is designed to lower the flowing bottomhole pressures further, thereby changing the flowing conditions, particularly the terminal decline. Application of Decline Curve Models with Changing Reservoir and/or Hydraulic Fracture Properties: The Arps models also assume implicitly that reservoir and fracture properties will remain constant during the boundary-dominated flow period. Stress-dependent reservoir properties are a common characteristic of tight gas sands, especially those that are initially abnormally overpressured. It is also not uncommon to have changes in fracture properties (i.e., effective fracture conductivity and/or length) over time. Common causes include twophase flow, ineffective cleanup of fracturing fluids, and loss of effective fracture conductivity due to proppant crushing or embedment. Reserves Validation Stage: Computing Volumetric Gas-in-Place (VGIP) In this section, we discuss the first reserves validation stage illustrated in Fig. 1. The overall objective of this stage is to estimate the volumetric gas-in-place (VGIP), i.e. the size, shape and content of the hydrocarbon container. Critical to this estimate are measurements of rock, pore and fluid properties and their distribution (both lateral and vertical). The two most common industry methods for estimating VGIP are volumetric mapping and conventional material balance methods. 5,22,23 Conventional material balance techniques are theoretically applicable to tight gas sands, but often cannot be implemented practically. The primary limitation is the availability of accurate, stabilized reservoir pressure histories. Wells completed in tight gas sands usually require very long shut-in times to reach stabilized pressures representative of the contacted pore volume, and most operators are reluctant to shut in a well for extended periods, particularly under favorable natural gas pricing scenarios. Consequently, accurate, reservoir pressure histories are often not available or may be inaccurate because of inadequate shut in times. Moreover, material balance methods require both pressure and production histories from boundary-dominated flow 22 which may not occur for many years. Substantial errors 24,25 will occur when conventional material balance methods are applied in tight gas sands with inaccurate pressure histories. Because of the restrictions and problems with application of conventional material balance models in tight gas sands, our methodology focuses on volumetric mapping techniques for estimating VGIP. More specifically, we address data acquisition and evaluation programs for constructing the requisite maps for estimating VGIP. Depending on the quantity and quality of data available, the VGIP evaluation can be made on any scale, i.e., either on a well, pattern, or field basis. Static VGIP estimates are often made early in the field development and production stage, but should be adjusted throughout the entire reservoir life cycle as more wells are drilled and more data become available. Accurate quantification of VGIP requires an integrated data acquisition and evaluation program that addresses all elements of the hydrocarbon system, i.e., a description of the rock, pores and fluids. 26 Major elements of this validation stage (Fig. 1) include monitoring of static reservoir pressures as well as acquisition and evaluation of both core and fluid data. A comprehensive well logging program is also very important since it is useful for integrating and up scaling smaller, porescale rock and fluid data into larger, reservoir-scale storage and flow capacity profiles. Static Reservoir Pressures. Initial reservoir pressures are required for correctly computing original VGIP and are critical for identifying and monitoring the various flow periods during field development. We recommend pressure measurements be made with bottomhole gauges; however, surface measurements may be adequate if obtained with static gradient tests to determine the type and quantity of liquids in the wellbore. We also recommend that static reservoir pressures be measured routinely as part of a field monitoring and surveillance program, especially in conjunction with infill drilling activities. When wells are completed in more than one layer, we also suggest that individual layer pressures be measured at the initial completion phase, particularly in key infill, field delineation and step-out wells. Core Acquisition Program. We recommend core samples be taken early in the field development stage, especially in key development and field delineation wells. Large-diameter, conventional whole core which we prefer to sidewall coring should ideally be obtained throughout the entire vertical section, including both reservoir and non-reservoir rock. Complete vertical sections are used for interpreting genetic units into depositional sequence and predicting depositional environment and architecture. Core data will also

6 6 J.A. Rushing, K.E. Newsham, A.D. Perego, J.T.Comisky, and T.A. Blasingame SPE help develop an understanding of reservoir geometry, continuity and distribution of rock types and properties. Core-derived connate water saturation measurements, especially in tight gas sands, are most accurate when nativestate or preserved core is used, so the coring program should be designed to minimize flushing and displacing connate water by the drilling fluids. We have observed significant mud invasion profiles caused by fluid imbibition from both waterand oil-base muds in tight gas sands. As a result, we recommend using either an oil-base mud or a low fluid-loss, ph-neutral, low-surfactant-energy water-base mud for coring operations. Alternatively, we may use various radioactive tracers in the mud which allows us to correct the extracted water volume back to native state concentration. Core Evaluation Program. Table 2 summarizes our recommended core analysis program which is designed to describe and quantify those properties that affect gas and water storage. The program is divided into three fundamental categories descriptions of rocks, pores, and rock-pore-fluid properties. Description of Rocks. The rock description program begins with a physical core description including interpretations of sedimentary structures, bedding morphology and textural characteristics. All of these attributes can be very helpful in classifying sequences of genetic depositional units which not only provide insight into depositional environment, but also provide bounding conditions on sand body geometry and continuity, thus helping describe dimensions of the original sand deposits. Petrophysical descriptions of sediment source, grain texture, composition, clay type and mineralogy are important since they not only affect rock storage and flow properties, but also help us to develop an understanding of the type and extent of diagenesis Diagenesis defined as any physical or chemical process causing changes in initial rock properties is the principal cause of reductions in both permeability and porosity subsequent to deposition. The primary diagenetic processes observed in low-permeability sands are mechanical and chemical compaction, quartz and other mineral cementation, mineral dissolution and clay genesis. Description of Pores. Detailed description and quantification of a rock s pore system are also essential for understanding fluid storage characteristics as well as the vertical distribution of fluids within the reservoir. Included in these properties are pore size distribution and geometry as well as porosity. Highpressure mercury injection is an effective technique to quantify pore geometry, particularly the size and distribution of pore bodies and throats Porosity is affected by the textural properties associated with primary depositional processes and by all subsequent diagenetic processes. Porosity may be classified into two fundamental types effective and total. Effective porosity quantifies only that pore volume that is connected, while total porosity is a measure of all pore volumes, regardless of their connectivity. 26 Tight gas sands may exhibit several types of effective porosity, including both primary and secondary. Primary inter-granular porosity exists as pore spaces between the rock framework grains, while secondary porosity is often caused by dissolution of soluble mineral grains. Microporosity, which is typically associated with clays and shales, can also be formed by isolation of pores from cement overgrowths. Fracture porosity is created from micro fractures in rock materials. Petrographic analyses and imaging of thin sections is the best method to estimate their relative percentages and contributions. Table 2 Summary of recommended core evaluation programs for estimating VGIP in tight gas sands General Category Physical Core Description Petrophysical Description Rock Diagenesis Pore Size, Geometry & Distribution Total & Effective Porosity Absolute Permeability Stress- Dependent Properties Core- Derived Water Saturation Gas-Water Relative Permeability Gas-Water Capillary Pressures Electrical Properties Measured Parameters Description of Rocks Sedimentary structures, bedding morphology, faults and fractures, texture and fabric Sediment source, framework, texture, composition, clay types, & mineralogy Mechanical & chemical compaction, quartz cementation, mineral dissolution, clay genesis Description of Pores Quantify pore geometry, size & distribution of pore bodies & throats Primary & secondary inter-granular, detrital matrix, microporosity, & grain fracture porosity Klinkenberg- Corrected Absolute Permeability Stress-dependent absolute permeability & effective porosity Measurement Technique Physical description of slab, acoustic or resistivity imaging Thin sections point counts, x-ray diffraction (XRD) & scanning electron microscopy (SEM) Thin section epifluorescence and cathodoluminescence, scanning electron microscopy (SEM) High-pressure mercury injection capillary pressure (MICP), SEM Gas expansion (Boyles Law) methods, NMR, MICP Multi-point steady-state Klinkenberg technique Absolute permeability: Multi-point steady-state Klinkenberg technique Effective porosity: Gas expansion methods Description of Rock-Pore-Fluid Properties Fluid Saturations Effective permeability to gas and water Capillary pressures vs. water saturation Formation Resistivity factor, Resistivity index, saturation exponent, cementation factor Dean Stark extraction technique Incremental phase trapping; Klinkenbergcorrections Hybrid vapor desorption with high-pressure porous plate or highspeed centrifuge 2-electrode or 4-eletrode for homogeneous and heterogeneous rock, respectively

7 SPE Beyond Decline Curves: Life-Cycle Reserves Appraisal Using an Integrated Work-Flow Process for Tight Gas Sands 7 Although not required for computing VGIP, absolute permeability is an important property for identifying hydraulic rock types and determining reservoir flow capacity. 26 Hydraulic rock types are key elements for identifying net sand thickness or pay, which is required for estimating recoverable VGIP. Absolute permeability is affected by basic pore-scale rock properties, such as grain size, packing, grain angularity, composition, and diagenesis. Permeability can be measured with either a steady-state or unsteady-state permeameter; however, we recommend a steady-state technique for rocks with micro-darcy permeability. 32 We also recommend that all permeabilities be corrected for gas slippage or Klinkenberg effects. 33 Tight gas sands also typically display stress-dependent porosity and permeability characteristics. 34 Reductions in porosity, however, are usually much less significant than in permeability. The magnitude of loss is also a function of original net mean stress conditions and the expected abandonment pressure, so rock properties should be evaluated over the entire range of expected net stress conditions, i.e. covering the complete production and pressure decline history. Description of Rock-Pore-Fluid Properties. The last phase of our recommended core evaluation program focuses on those properties that depend on interactions among the rock surfaces, pores, and fluids occupying the pores. Included in these properties are core-derived water saturations, gas-water effective and relative permeabilities, gas-water capillary pressures, and electrical properties. Core-derived water saturations are typically measured in the laboratory with the Dean Stark extraction technique 35,36 which uses a hydrocarbon solvent to leach connate water out of the connected pore space. Consequently, Dean Stark extraction is best performed on native-state, center-cut plugs where core filtrate invasion is minimal. Although this method is very effective, clay dehydration (caused by the high temperatures required to reach the solvent s boiling point) is a possibility. Dehydration can alter the clay fabric and induce desiccation cracks which are manifested by measured permeabilities that are much larger than the intrinsic permeability. Although not necessary for computing VGIP, effective and relative permeabilities are required for reservoir modeling and well performance analysis to estimate CGIP. Effective permeability is a function of the rock s wettability, pore geometry, and saturation of both wetting and non-wetting fluid phases. 35 Some tight gas sands may have little or no mobile water phase in the reservoir, so we have utilized a technique called incremental phase trapping 37 in which the core plug water saturations are increased incrementally. At each saturation, the gas effective permeability is measured by flowing vapor-saturated gas through the system. To eliminate adverse rock-fluid interactions, we should also use either connate water or synthetic water with similar properties to saturate the core. Finally, we should measure Klinkenbergcorrected effective permeabilities. 38 Gas-water capillary pressure characteristics affect the vertical distribution of fluids as well as the producing characteristics in tight gas sands. A number of techniques have been developed for measuring capillary pressure characteristics in conventional reservoirs; however, limitations on the maximum measurable pressures preclude their applicability in tight gas sands. Consequently, we recommend a combination of vapor desorption with either high-pressure porous plate or highspeed centrifuge. 39,40 We also strongly recommend that capillary pressure characteristics be measured using actual reservoir fluids (i.e., gas and water) rather than mercury so that we can more accurately incorporate rock wettability. 39 Electrical properties provide an important link between corederived, pore-scale petrophysical properties and log-derived, meso-scale reservoir properties. Electrical properties are measures of a rock s ability to conduct electric current. Conductance is due primarily to the movement of dissolved ions in the brine saturating the rock pores and varies directly with ion concentration. 35 Temperature, porosity, pore geometry, and rock composition also affect the rock conductance. Electrical properties are usually presented in terms of a formation resistivity factor, resistivity index, tortuosity factor, saturation exponent, and cementation factor. The ions in clay-bound waters may also act as separate conductors. This excess conductance must be incorporated into water saturation models, particularly when calibrating the log response to core-derived parameters. Excess conductance is commonly estimated using the C o /C w method. 36 The conductance of a core sample saturated with fluids of varying salinity is plotted against the conductivity of the saturating fluid. From this plot, we can graphically compute a pseudo water saturation exponent and pseudo cementation factor. These pseudo parameters allow us to use Archie s 41 model and account for excess conductance. Acquisition of Fluid Samples. All fluids gas, condensate, and water samples should be taken for laboratory analysis and phase behavior studies. Except for retrograde gas condensates, samples obtained from surface separators are usually sufficient for dry or wet gas reservoirs. 42 A description of the hydrocarbon components should include the fluid type, phase behavior and any associated nonhydrocarbon contaminants. McCain 42 has shown that reservoir fluid type is confirmed most accurately by laboratory measurements; however, in lieu of laboratory analyses, he has provided rules of thumb to identify fluid type from initial producing properties. Consequently, the fluid volumes, properties, and compositions of all produced phases should be measured and documented early and consistently. Fluids Evaluation Program. A description of the reservoir fluids should not only qualify but also quantify the fluids occupying the pore spaces. These descriptions should include hydrocarbons, non-hydrocarbon contaminants (e.g., CO 2, N 2, H 2 S, water vapor, etc.), and connate water. Basic gas properties, such as composition, z-factor, compressibility, and viscosity, can be estimated from industry-standard correlations; however, care should be taken in highpressure/high-temperature (HP/HT) gas reservoirs in which the environmental conditions exceed pressure and temperature ranges used to develop the correlations. Non-hydrocarbon components may affect the gas properties, so we should also measure these components. Industrystandard correlations are also available to correct gas

8 8 J.A. Rushing, K.E. Newsham, A.D. Perego, J.T.Comisky, and T.A. Blasingame SPE properties for most common contaminants Note that the source of water produced from tight gas sand reservoirs, especially from reservoirs at HP/HT conditions, may not be from a mobile liquid phase in the reservoir, but rather may be condensed water vapor that is initially dissolved in the reservoir gas phase. As a result, we recommend that fluid samples be taken and analyzed periodically. Measured water salinities from produced waters should be compared to that determined from the connate water saturation. Well Logging Acquisition Program. We recommend that a comprehensive well logging program be implemented for the entire field development. Most standard logging packages include gamma ray, spontaneous potential, resistivity, photoelectric and porosity (neutron and density) logs. Depending on the situation, we recommend other logs be considered. For example, the addition of a dipole sonic log provides more flexibility to identify complex lithologies and minerals and also allow us to estimate dynamic in-situ elastic and mechanical rock properties including rock strength, stress regime and principle stress magnitudes. Borehole imaging logs, either resistivity- or sonic-based tools, should be considered to identify natural fractures, quantify bedforms and orientation, and stress and hydraulic fracture orientation. Where a relationship between pore body and pore throat size exist, nuclear magnetic resonance logs may be useful for defining porosity size distributions and absolute permeability. Through casing production logs are very helpful for defining time-lapse changes in the gas in-flow profiles. Temperature, gradiometer, spinners, and cavitation indicators should be run multiple times over a well s production history. Well Log Analysis. The objective of well log analysis is to upscale the pore scale descriptions to continuous vertical distributions of porosity, permeability, fluid saturations and net pay. Since physical measurements of the rocks, pores, and fluids provide ground truth references, the log response should be calibrated with the core data and incorporated into the log analysis. Rocks and Pores. Knowledge of the rock composition, including clay type and distribution within the pore structure, helps calibrate nuclear responses to matrix, clay and pore volumes and aids in the selection of an appropriate porositysaturation model. Compositional analysis of produced fluids, including both hydrocarbons and water, provides estimates of fluid density, salinity, and compressibility. For example, gas properties are necessary inputs to model the light hydrocarbon fluid effects on the log responses and to obtain accurate porosity estimates. Rock-Pores-Fluids. Water saturation is a critical parameter for estimating gas-in-place and reserves. One critical parameter required to estimate water saturations from the log response is the connate water resistivity, R w. We have identified four methods to estimate R w. Chemical analysis of the produced waters is the most direct technique to estimate R w. If, however, there is no mobile water phase in the reservoir or if the produced water has been contaminated with drilling and completion fluids, then we must rely on other techniques, e.g., empirical log-based methods including SPderived, apparent water resistivity or Pickett plot techniques. 46 We also recommend commutation or residual salts analysis 47 which extracts or leaches connate water and the associated salts from preserved, native-state core samples using an ultrapure, de-ionized water. Salt concentration and composition in the leachate are measured using an atomic absorption or mass spectrometer technique, and salinity is estimated from material balance calculations. Fluid inclusion micro-themometry uses thin sections from preserved core samples to measure the temperature at which fluid inclusions melt. The melting temperature is directly related to the connate water salinity. Measured electric properties, such as saturation exponent and cementation factor, help to define the variability of the Archie parameters. Practical Considerations in Volumetric Mapping. As we stated previously, the overall objective of the first reserves validation stage is to estimate VGIP, or more specifically, to estimate the size, shape, and content of the hydrocarbon container. Estimates of VGIP can be made on any scale, i.e., either on a well, pattern, or field basis. Moreover, the accuracy of VGIP estimates from mapping depends not only on the type, quantity, and quality of data but also on the field development stage. Well control and the overall field well spacing compared to the size of the sand deposits and the hydrocarbon accumulation are important considerations in volumetric mapping. Early in the field development phase when few wells have been drilled, we may have to incorporate seismic data to capture large-scale geologic attributes in our volumetric maps. During the intermediate and final stages of the field development, we should be able to incorporate more local geologic aspects and refine our maps and subsequent VGIP estimates. Knowledge of regional geology will also help in defining the size and geometry of the hydrocarbon container or field. As we stated earlier, interpretations of sequences of genetic units from the core description program may not only provide insight into depositional environment, but also provide bounding conditions on sand body geometry and continuity, thus helping describe dimensions of the original sand deposits. Depending on the type of data available, we may have to rely upon our field analog for estimating reasonable dimensions and orientation of the sand deposits. As more wells are drilled and well control is established, we should be able to adjust the initial, large-scale volumetric map to smaller-scale maps that capture both local geologic features, reservoir continuity, and sand-body heterogeneities. Understanding relationships between structure, fluid contacts and distribution of fluids, and variations in pore volume properties requires at least a structure map and an area or volume map. 23,51 Since the reservoir is a three-dimensional structure, vertical cross-sections and fence diagrams may also be required to better understand sand continuity. Fundamental to estimating VGIP is a two-dimensional map of the hydrocarbon-filled pore volume, so we recommend using maps of the hydrocarbon pore volume, i.e. the product of effective porosity, net pay and gas saturation (φhs g ). This type of map allows us to incorporate both lateral and vertical variations in porosity and water saturation. VGIP is computed by planimetering contours of φhs g. 23,49

9 SPE Beyond Decline Curves: Life-Cycle Reserves Appraisal Using an Integrated Work-Flow Process for Tight Gas Sands 9 Net pay which is defined as the net reservoir interval that contains producible hydrocarbons is estimated by applying a cutoff to the gross sand thickness. 52 Effective porosity has historically been used to generate cutoffs in tight gas sands, but we recommend both storage and flow capacities to identify net sand thickness. Diagenesis typically reduces the permeability of the pore throats connecting the pores much more than the pores, so we have often observed permeability differences of several orders of magnitude for the same effective porosity. Similarly, Worthington 53 has suggested cutoffs should not only be dynamically conditioned to incorporate both flow and storage capacities, but should be fit-for-purpose, i.e., the cutoff should be compatible with the reservoir depletion mechanism as well as the type and scale of data. Therefore, we recommend using a combination of hydraulic rock typing combined with effective porosity, effective permeability, and water saturation delimiters to estimate net pay. 26,54 Volumetric gas-in-place may be computed deterministically or stochastically. Deterministic estimates are typically obtained by planimetering maps of net hydrocarbon thickness. Stochastic models use principles of statistics and probability to distribute geological structures and/or rock properties in order to provide estimates of uncertainties on these distributions. 52 In general, applications of stochastic models include capturing distributions of both large-scale geological and small-scale petrophysical properties; incorporating all reservoir heterogeneities; quantifying all possibilities of sand continuity between wells; and inclusion of rock property anisotropies and their distributions between wells. The principal advantage of a stochastic evaluation is the ability to provide a range of VGIP estimates and to quantify their uncertainty. Reserves Validation Stage: Computing Contacted Gas-in-Place (CGIP) In this section, we discuss the second reserves validation stage illustrated in Fig. 1. The objective of this stage is to estimate the contacted gas-in-place (CGIP) volume. We define the CGIP as the fraction of VGIP that is in pressure communication with or connected to the producing wells. Quantification of this volume essentially describes reservoir continuity and provides estimates of the recoverable VGIP or reserves. Moreover, comparison of the total CGIP volume with the estimated VGIP volume quantifies recovery efficiency and may indicate infill drilling potential, especially if the CGIP is significantly less than VGIP during the intermediate or late field development stages (Table 1). As we discussed previously, CGIP is a dynamic volume that changes over time depending on field development maturity, well spacing, stimulation effectiveness etc. The major elements of this reserves validation stage include a well surveillance and monitoring program, well performance evaluations, and reservoir simulation. Similar to the VGIP calculations, the CGIP can be evaluated either on a well, pattern, or full field basis depending on the quantity and quality of data. However, several of the well performance evaluation techniques are best applied on a single-well basis, while reservoir simulation can be done on any scale. Well Surveillance and Monitoring Program. The first element of our work-flow process for estimating CGIP is a reservoir monitoring and surveillance program. The objective of this phase is to continuously monitor well production performance and to obtain the requisite data to quantify any anomalous well production decline behavior. These data are also very important as input for various types of reservoir models used to quantify CGIP. Fluid Production Rates. Although produced hydrocarbon fluid volumes are normally recorded daily as part of routine field operations, we also recommend that water volumes be recorded. Monitoring produced water volumes may help identify and solve potential reservoir or operational problems. For example, excessive produced water volumes may indicate communication with overlying or underlying zones either because of ineffective cementing or from hydraulic fracture growth into these wet zones. On the other hand, anomalously low produced water volumes in high-pressure/hightemperature gas reservoirs may be caused by condensation of water vapor dissolved in the hydrocarbon gas. Comparison of compositional analyses from produced and in-situ connate waters may also help identify the source of the produced water volumes. Flowing Well Pressures. We also suggest that the flowing pressures associated with the fluid production histories be recorded. Continuously monitored bottomhole flowing pressures offer the best data source for identifying well problems and optimizing production; however, economic considerations often preclude implementing this type of program. A good alternative is acquisition of accurate wellhead flowing pressures. If there is little or no liquid accumulation in the bottom of the well, bottomhole flowing pressures may be estimated accurately from wellhead values. A production logging program should also be implemented if there is significant liquid production and/or if the productivity is affected by liquid loading. Ideally, production logs should be measured not only during the early producing years, but also frequently during the well s entire productive life. If wells are completed in several major sands separated by large intervals, we also strongly recommend that a production logging program be implemented to routinely measure bottomhole flowing pressures and gas in-flow by zone. Production logs may also help identify operational problems or highlight intervals of insufficient reservoir stimulation. Flowing pressures are also vital input for both well performance analysis and reservoir simulation. Regardless of the technique employed, problems with production data quantity and/or quality e.g., incomplete or infrequent data sampling; poor quality data; and/or erroneous data affect the accuracy of the analyses. This impact is especially problematic for accurate assessments of both fracture and reservoir properties using transient data which are typically changing frequently and rapidly during early-time flow periods. Monitoring Production Decline Behavior. Although ratetime plots are normally used just to estimate gas reserves, these plots may also be excellent diagnostic tools. More specifically, monitoring changes in the computed and expected