2009-23168.pdf

11
SPWLA 50 th Annual Logging Symposium, June 21-24, 2009 1 A ROCK MODEL FOR SHALE GAS AND ITS APPLICATION USING MAGNETIC RESONANCE AND CONVENTIONAL LWD LOGS Daniel F. Coope, Terrence H. Quinn, Elton Frost Jr. and Michael J. Manning, Baker Hughes, Inc. Copyright 2009, held jointly by the Society of Petrophysicists and Well Log Analysts (SPWLA) and the submitting authors. This paper was prepared for presentation at the SPWLA 50 th Annual Logging Symposium held in The Woodlands, Texas, United States, June 21-24, 2009. ABSTRACT Petrophysical evaluation of gas in shales is established as a largely empirical practice based simply upon a shale-plus-kerogen-plus-fluid model. Pragmatism has stemmed from the practical considerations of a limited data set as much as from an undeveloped physical concept. Until recently logging measurements collected in shales have been the minimal set; further, the physical concepts proven in our familiar Archie rocks are both conceptually and practically unappealing in source rocks. Recently more numerous and specialized measurements have been run in these rocks, including natural gamma and pulsed neutron spectroscopy logs, images and magnetic resonance. At present, a primary application of both wireline and LWD in resource plays is the use of image logs to optimize borehole placement and stability and to maximize hydraulic fracture effectiveness. With more information available from these rocks, a more detailed petrophysical rock model should and has been considered by various authors (Fertl et al. 1988). We extend evaluation in two areas—evaluation of the rock in terms of its mass constituents and pore space, and detailed application of the particular features of magnetic resonance logs in shale. The objective of the approach is not only to determine total organic carbon but to have a quantitative description of kerogen properties in situ. Magnetic resonance is particularly appealing in unconventional resource rocks since it delivers, unlike all other log methods, a model-independent determination of porosity. In addition, the relaxation distribution can insinuate shale type and “richness”, especially interesting for our approach. The incentive for our work is to evaluate the resource in real time, enabling better placement and completion of horizontal wellbores. At present, total organic carbon maturation and kerogen type are determined from combined laboratory measurements and experience. The more thoroughly we evaluate organic carbon in situ, using real-time log measurements, the more drilling and completion costs will decrease. We demonstrate our method using LWD logs in the Woodford shale. Our data set encompasses large-diameter core and wireline and LWD measurements in several source rock shales of varying geological and petrophysical properties. Core analysis of shale constituents and organic carbon content is compared to the interpretation model results. INTRODUCTION Resource rock models need to relate log measurements to the important parameter, total organic content (TOC), and at best be able to predict the level of (kerogen) maturation (LOM) and kerogen type. Conceptually, the rock physical model used for our shale gas evaluation is the same as earlier models (e.g,. Passey et al. 1990). Figure 1 shows the model. It describes the source shale as a combination of shale, organic material and fluid-filled pore space. We are specific in describing the source rock as gas-prone shale, but other matrix, such as mudstone, can be similarly conceptualized. In the case of the matrix containing elemental or inorganic carbon, one can account for that explicitly in the evaluation model. As the organic matter matures, gas is ejected from the solid bitumen into the pore space. The gas displaces the connate water, leaving the pore space filled with gas and irreducible water. This model is vague in differentiating free, adsorbed and absorbed gas. It has been used as the concept behind the ‘ΔLogR’ (Passey et al. 1990) evaluation technique, which quantifies TOC, requiring a measure of LOM. ΔLogR Model—The ΔLogR method reasons that in rocks containing only brine in the pore space and low organic content, the bulk density and formation resistivity logs will, when normalized, overlay, since both logs respond to porosity. A similar argument can be made if density is replaced with either the acoustic slowness or neutron porosity. But in rocks with organic

Upload: aliniski123

Post on 09-Feb-2016

20 views

Category:

Documents


0 download

DESCRIPTION

This is TOC and LOM estimation. In addition to the petrophysical aspects

TRANSCRIPT

Page 1: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

1

A ROCK MODEL FOR SHALE GAS AND ITS APPLICATION USING MAGNETIC RESONANCE AND CONVENTIONAL LWD LOGS

Daniel F. Coope, Terrence H. Quinn, Elton Frost Jr. and Michael J. Manning, Baker Hughes, Inc.

Copyright 2009, held jointly by the Society of Petrophysicists and Well Log Analysts (SPWLA) and the submitting authors. This paper was prepared for presentation at the SPWLA 50th Annual Logging Symposium held in The Woodlands, Texas, United States, June 21-24, 2009. ABSTRACT Petrophysical evaluation of gas in shales is established as a largely empirical practice based simply upon a shale-plus-kerogen-plus-fluid model. Pragmatism has stemmed from the practical considerations of a limited data set as much as from an undeveloped physical concept. Until recently logging measurements collected in shales have been the minimal set; further, the physical concepts proven in our familiar Archie rocks are both conceptually and practically unappealing in source rocks. Recently more numerous and specialized measurements have been run in these rocks, including natural gamma and pulsed neutron spectroscopy logs, images and magnetic resonance. At present, a primary application of both wireline and LWD in resource plays is the use of image logs to optimize borehole placement and stability and to maximize hydraulic fracture effectiveness. With more information available from these rocks, a more detailed petrophysical rock model should and has been considered by various authors (Fertl et al. 1988). We extend evaluation in two areas—evaluation of the rock in terms of its mass constituents and pore space, and detailed application of the particular features of magnetic resonance logs in shale. The objective of the approach is not only to determine total organic carbon but to have a quantitative description of kerogen properties in situ. Magnetic resonance is particularly appealing in unconventional resource rocks since it delivers, unlike all other log methods, a model-independent determination of porosity. In addition, the relaxation distribution can insinuate shale type and “richness”, especially interesting for our approach. The incentive for our work is to evaluate the resource in real time, enabling better placement and completion of horizontal wellbores. At present, total organic carbon maturation and kerogen type are determined from

combined laboratory measurements and experience. The more thoroughly we evaluate organic carbon in situ, using real-time log measurements, the more drilling and completion costs will decrease. We demonstrate our method using LWD logs in the Woodford shale. Our data set encompasses large-diameter core and wireline and LWD measurements in several source rock shales of varying geological and petrophysical properties. Core analysis of shale constituents and organic carbon content is compared to the interpretation model results. INTRODUCTION Resource rock models need to relate log measurements to the important parameter, total organic content (TOC), and at best be able to predict the level of (kerogen) maturation (LOM) and kerogen type. Conceptually, the rock physical model used for our shale gas evaluation is the same as earlier models (e.g,. Passey et al. 1990). Figure 1 shows the model. It describes the source shale as a combination of shale, organic material and fluid-filled pore space. We are specific in describing the source rock as gas-prone shale, but other matrix, such as mudstone, can be similarly conceptualized. In the case of the matrix containing elemental or inorganic carbon, one can account for that explicitly in the evaluation model. As the organic matter matures, gas is ejected from the solid bitumen into the pore space. The gas displaces the connate water, leaving the pore space filled with gas and irreducible water. This model is vague in differentiating free, adsorbed and absorbed gas. It has been used as the concept behind the ‘ΔLogR’ (Passey et al. 1990) evaluation technique, which quantifies TOC, requiring a measure of LOM. ΔLogR Model—The ΔLogR method reasons that in rocks containing only brine in the pore space and low organic content, the bulk density and formation resistivity logs will, when normalized, overlay, since both logs respond to porosity. A similar argument can be made if density is replaced with either the acoustic slowness or neutron porosity. But in rocks with organic

Page 2: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

2

carbon, the two logs will separate. Bulk density is reduced by the presence of the lighter organic or inorganic carbon, suggesting increased porosity over the true porosity. Further, if the organic matter has expelled hydrocarbon into the pore space, the resistivity log will read higher, proportional to the organic maturation. Displayed using a standard log presentation, the porosity log swings to the left, and the resistivity log moves to the right, anti-correlating. (Note that these qualitative observations have been used from the earliest days of log analysis in reservoir rocks.) True porosity changes in the source rock will result in the porosity log and resistivity log moving in correlation. The calibrated separation of these two logs yields an estimate of TOC.

Immature Organic Matter

SolidPore

Space

Water Filled

(Pore Space)

Solid

Mature Organic Matter

SolidPore

Space

Water Filled

(Pore Space)

Solid

Shale “Matrix” Shale “Matrix”G

as

Immature Organic Matter

SolidPore

Space

Water Filled

(Pore Space)

Solid

Mature Organic Matter

SolidPore

Space

Water Filled

(Pore Space)

Solid

Shale “Matrix” Shale “Matrix”G

as

Fig. 1. Depiction of the rock plus pore space model for application to shale gas evaluation. The solid matter consists of “shale” and organic matter. As the organic matter matures, gas is expelled from the solid into the pore space, driving out some of the connate water. Multi-mineral Model—The ΔLogR model is empirical, based on pragmatism and proven successful. We will term the TOC evaluation method using an analytical solution to the rock model the “multi-mineral (shale gas) model”. The term analytical is used here in the strict sense of defining the rock by its particular constituents and solving for fractions of each. This approach defines the rock by its detailed constituents. Carbon can be explicitly assigned to particular minerals containing carbon (e.g., siderite, elemental carbon, and organic carbon). The model quantifies TOC as the carbon that is “left over”; that is, the carbon with the elemental and non-organic carbon “netted-out”. As an aside, evaluation of resource rocks focuses on the shale or mudstone. This is the rock that is usually “netted out” in reservoir rock evaluation. In reservoir evaluation, we specify matrix and shale, where shale is sometimes assumed to have no porosity or permeability. Here, we sometimes use the term matrix or shale matrix to refer to non-reservoir shale. This requires us to quantify the shale matrix parameters, such as matrix density, more carefully than we are accustomed to. We use crossplots and, in particular,

magnetic resonance (MR) logs, to obtain shale parameters. The solution technique employs the analytical methods of contemporary petrophysics, whence we relate independent measurements to the unknowns of the shale gas model above, where all the significant constituents of the rock are explicitly parameterized. Jacobi et al. and Gladkikh et al. published a multi-mineral solution for the Woodford and Barnett shales. Seventeen minerals and organic carbon were included. This solution was enabled by a full complement of modern logs, including geochemical (Pemper et al. 2006) and MR. Figure 2 depicts the multi-mineral model specified in that analysis. Fig. 2. Extension of the shale gas rock plus pore space model. The shale is parameterized by the properties of its constituent parts. Beyond these many high-tech measurements, additional information is needed. An “expert system”, incorporating facies identification and a laboratory-based catalog of mineral parameters, was developed. Results are displayed in the paper as constituent weight percent (wt%)—see Appendix II. TOC is computed as the carbon wt% after elimination of carbon contributions from calcite, dolomite and siderite.

LIMITED MINERAL MODEL

The conceptual model for gas-bearing shale (Figure 1) has worked well for both the ΔLogR and multi-mineral TOC evaluation approaches. However both ΔLogR and the multi-mineral approaches have limitations for evaluating TOC. ΔLogR is based on empiricism, not necessarily a bad thing, but inherently not analytically adaptable to varied matrix properties or other log measurements. The multi-mineral approach is limited to evaluations that have a complete suite of “high tech” logs and preferably cuttings and core data.

Mature Organic Matter

PoreSpace

Water Filled

(Pore Space)

Solid

Gas

Quartz Dolomite Calcite Salt

Illite Smectite Kaolinite Chlorite

Glauconite Hematite Siderite

K-feldspar Pyrite Apatite Zeolites

Plagioclase Anhydrite

Mature Organic Matter

PoreSpace

Water Filled

(Pore Space)

Solid

Gas

Quartz Dolomite Calcite Salt

Illite Smectite Kaolinite Chlorite

Glauconite Hematite Siderite

K-feldspar Pyrite Apatite Zeolites

Plagioclase Anhydrite

Quartz Dolomite Calcite Salt

Illite Smectite Kaolinite Chlorite

Glauconite Hematite Siderite

K-feldspar Pyrite Apatite Zeolites

Plagioclase Anhydrite

Page 3: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

3

Limited Mineral Model (LMM)—Our evaluation approach is motivated by the need to determine TOC from a few measurements, as is usual for LWD conveyed logging suites. Accepting the rock model, Figure 1, with the adaptations of Figure 3, we solve for TC using an analytical method. We characterize the solid as “shale” and carbon, calling this approach the “limited mineral” model. It can be extended to contain a light or heavy mineral that may, if ignored, bias the result. Fig. 3. The Limited Mineral Model treats the shale as an entity with consideration for a single heavy or light mineral. The differences between our method and the “multi-mineral” model are:

1. We do not solve for TOC, but rather TC. We will infer TOC from MR T2 (in a later section).

2. MR porosity is included in the set of independent “knowns” to solve for TC.

3. MR array data are a key input:

a. Effective porosity (MPHE) and clay-bound water (CBW), are used as “soft” independent knowns;

b. The T2 distribution is recognized as a possible method of determining shale and kerogen type.

4. We solve for TC using an error minimization

technique, including uncertainty. The algorithm compares the theoretical response for each log with the measured log value, minimizing the differences and explicitly presenting the degree of uncertainty. This allows us to include the effect of proxy heavy or light minerals and evaluate the validity of crossplot parameters (e.g., neutron matrix response).

Carbon can occur as elemental and molecular constituents of the resource rock. Often it is easily

distinguished from organic carbon by its petrophysical properties; for instance limey shales have much higher grain density. Coal is removed using a gamma ray cutoff. Our solution may contain a minor non-organic carbon component, but often prior knowledge eliminates this concern. However, we present our result as TC, to remind ourselves that assumptions are needed to differentiate TOC and TC. Later we also point out that the MR T2 shale and irreducible water (BVI) response may help us identify TOC and even LOM. Below, the limited shale gas model is demonstrated using logs and core results from the Baker Hughes Experimental Test Area (BETA) N5 test well drilled into the Woodford shales about 20 miles south of Tulsa, Oklahoma, USA. Measured were MR (MPHS and T2), bulk density (RHOB) and neutron porosity (NPHI). (Acoustic porosity could also be used, but was not in this case). The model has seven unknowns:

ϕmatrixcarbonshalewatercarbonshale N ,f ,f ,ρ ,ρ ,ρ φ,

where ρ is density, ϕ is porosity, f is fraction, and

N matrixϕ is the neutron response in 100% solid matrix.

We gain a fourth relationship by recognizing material balance:

carbonshale f f φ1 ++=

ϕmatrixN is determined from neutron, bulk density and

MR crossplots. Finally, assuming constant values for

watercarbon ρ ρ and , we reduce the problem to four (4) relations to solve for four (4) unknowns (see Appendix I):

.N ,f ,ρ φ, matrixcarbonshaleϕ

The objective of our evaluation is to determine TC by wt%; TC weight is carboncarbon ρ f × . WOODFORD SHALE EXAMPLE We applied our evaluation methods to the Woodford shale formation. The silica-rich Woodford shale is known to have Type II (oil-generative) kerogen content with TOC adequate to produce oil and gas. Figure 4 shows the LWD log set from the BETA well, N5. Core (Figure 5) was collected in this and other test wells nearby. These log and core data are compared with the ΔLogR and LMM models.

Carbon

PoreSpace

Water Filled

(Pore Space)

Solid

Gas

Shale “Matrix”Light/Heavy Minerals

Carbon

PoreSpace

Water Filled

(Pore Space)

Solid

Gas

Shale “Matrix”Light/Heavy Minerals

Page 4: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

4

Fig. 4. Input log data used to solve for total carbon using the LMM with logs gamma ray, propagation resistivity (Rt), bulk density, neutron and MR porosity. Gamma ray defines the shale; note the very high API value typical of the Woodford. The high resistivity is indicative of organic content.

Fig. 5. Woodford shale core from BETA well N5. The core was cut in a deviated wellbore of about 55° inclination and is seen in a visible light photo. Breakages along bedding planes are caused by release of overburden pressure upon retrieval of the core. Within the black shale are occasional minor, silty or limey bedding, also observed by micro-electrical imaging logs, Fig. 9.

Woodford Shale Geology and Data Set—The BETA N5 well is located in the area of the now-depleted Glenpool Oil Field. Structurally, this area lies in the northeastern Central Oklahoma platform. Paleozoic bedding planes are mildly dipping in this area so that multiple boreholes with various wireline and LWD log data sets in the test area can be easily correlated. Drilled wells penetrate Paleozoic rocks from Pennsylvanian at the surface to Ordovician age at TD. During the Paleozoic, this area was on the distal flank of the Ouachita uplift (Cardott 2008). Strata of the Silurian and Early and Middle Devonian are absent in northeastern Oklahoma (Hinch and Derby 1998); there is a major regional unconformity separating Ordovician and late Devonian sediments. Immediately above the unconformity at BETA are the upper Devonian Misener sandstone, 2-ft thick, and then the 31-ft-thick Woodford black shale. Above the Woodford, Hinch and Derby describe a minor unconformity and a greenish-gray Kinderhook Shale of the St. Joe Formation. The most notable petrophysical properties of the Woodford shale at BETA are the very high natural gamma ray response on the order of 700 API units and very high organic carbon content. The lower section of the Woodford also has abundant pyrite nodules. Whole core appears to be dark black, shiny, with subtle laminar bedding surfaces as per the core photo, Figure 5. Bedding is also resolvable by micro-electrical imaging logs. According to both Comer et al. 1987 and Cardott 2008a, the Woodford Shale in the region south of Tulsa contains abundant oil-generating organic material of Type II kerogen with up to about 14% by weight TOC. Cardott 2008b cites one data point very near the BETA N5 site with a vitrinite reflectance (isoreflectance) value of 0.82%. Comer provides a more general mapping with regional values in the oil window range of 0.60 to 1.50% vitrinite reflectance. Comer gives a lower value of TOC for the region. The Glenpool field area has produced large volumes of oil since discovery in 1905, with most of the oil thought to originate in the Woodford, and a lesser volume possibly generated in the Fayetteville shale, several hundred feet above the Woodford in N5. Figure 4 presents the input log data: gamma ray, propagation resistivity (Rt), bulk density, neutron and MR porosity. Gamma ray defines the shale; note the very high API value typical of the Woodford. In the far right porosity track, the three nominal porosity logs are significantly different. The nature of the difference is a key to the success of our method. All three logs are porosity-dependant, but both neutron

Woodford Shale

Page 5: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

5

and density measurements are also dependant on matrix properties. Neutron porosity is influenced by slow neutron absorption in shale; bulk density is a function of mineral grain density. MR is a pure porosity measurement, independent of the matrix properties. These dependencies allow us to solve for TC considering the other matrix constituents. The equations relating log measurements to formation parameters are detailed in Appendix I. ΔLogR Applied in Woodford Shale—As a baseline for comparing methods, TOC was calculated using the ΔLogR method from both bulk density and neutron porosity. This method relies on the high resistivity value inferring the carbon is organic. The method requires a value for LOM; LOM = 8 was assumed in this case. Figure 6 shows the ΔLogR results.

Fig. 6. TOC calculated from ΔLogR method using both bulk density (red) and neutron porosity (blue). The red curve, based on density/Rt, is essentially the same as the neutron/Rt result in blue, suggesting about 15% TOC by weight. LMM in Woodford Shale—Application of our LMM assumed solid material composed of shale, carbon and a heavy mineral. The solution method minimizes the difference between the measured response inputs and theoretical responses calculated from the sensor response equations. The method accounts for a

prescribed uncertainty. Since we do not include resistivity in our method, we do not assume the carbon is organic, nor do we assume a level of maturation. Thus, TC is calculated. Figure 7 displays our result as the gray curve; the gray shading is interpreted as “excess,” non-organic carbon.

Fig. 7. Comparison of TOC calculated from the LMM and the ΔLogR methods (Fig. 6). The gray shaded area is the LMM result. LMM is either equal to or greater than the ΔLogR result. It is encouraging that the TC predicted by our LMM is bounded on the low side by TOC from the ΔLogR method. However, we will demonstrate that inclusion of “soft” MR parameters can significantly improve the calculation. Including MR Clay Water and Irreducible Water Porosity—We consider CBW and BVI as “soft knowns.” The determination of both is subjective, unless they can be verified independently from other information. These parameters allow us to over-determine the problem, reducing uncertainty; for instance “wet” shale volume, determined from conventional V-shale log analysis can be related to CBW to add another known. The method of error minimization is illustrated in Figure 8. For the porosity curves RHOB, NPHI, MPHE and BVI, Tracks 2 and 4, and the shale index, Track 5, we see the measured values closely overlaying theoretically predicted responses; the solid curves are the measured inputs, the dashed curves are the calculated predictions. These are calculated from the response equations corresponding to the volumetric

Page 6: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

6

solution, Track 1. Track 6 shows the minimized error function for this solution.

Fig. 8. The results of application of the error minimization calculation for the Woodford shale input. Tracks 2, 4 and 5 compare the input logs, solid curves, with the theoretical logs calculated from the respective sensor response functions and the volumetric result in Track 1. Track 6 shows the error function, used as a quality of result indicator. The LMM analytical solution is compared with the ΔLogR result in Figure 9 (found at end of paper), Track 6. We see the methods give essentially the same percent weight fractions of TOC, a significant improvement over the more simplified result of Figure 7. Although the LMM is not an improvement over the ΔLogR method in this case, it does provide a more satisfying general approach to calculating TOC. Foremost, it presents the calculated volumes within an integrated view of the rock-fluid system. It enables a simple method for evaluating the sensitivity of each parameter and quickly “flags” an unreasonable answer. Figure 9 displays the MR T2 distribution in the Woodford shale. This log was acquired using the “zero gradient” MagTrakTM LWD tool. It is unique among wireline and LWD MR logging measurements, having essentially no diffusion effect on T2. This is important for our application for two reasons. First, diffusion would have shifted the gas signal to the left of the spectrum, apparently increasing the irreducible water (BVI) and masking the presence of free gas. In our log, gas would appear in the far right of the spectrum; we see that we do not have measurable free gas, Figure 9, Track 5. Second, and important for our application, the absence of diffusion yields an unambiguous shale signature that we hope to cross reference to organic carbon content and maturation. This is considered in the next section.

APPLICATION OF THE T2 SPECTRUM

Without additional information, the LMM can only assume the computed carbon is organic, or better, seek descriptive core or offset data to remove non-organic carbon from TC. However our goal is to look for added information in the MR T2 distribution. We know that for some time researchers have attempted to relate the early (short relaxation time) T2 spectrum to shale type, with some success. Over 10 years ago, Prammer et al. showed that different clay minerals generally have signature T2 values. For instance, from that study chlorite would be expected to have a CBW T2 centered at 5 msec. Observing bitumen response from MR directly seems impossible, at least using current technology. Kerogen is characterized by its hydrocarbon index, which is in fact the property determined by MR. However, the relaxation times expected from hydrogen contained in organic molecules are too fast to be visible to our MR devices. If we are to classify organic shale content using MR, it will need to be through indirect identification. Close inspection of the T2 distribution, Figure 9 and Figure 10, reveals some differences in the BVI component in the Woodford shale; more BVI signal appears at the bottom. This is interesting since the entire section indicates TOC from both ΔLogR and the LMM. Yet at this point we can only speculate as to whether the difference could be due to different kerogen type or LOM. The signature is further confused considering that borehole washout also has the signature BVI increase. Note that just above the Woodford, the increased BVI is almost certainly due to borehole washout. The LWD electrical image in Track 2 confirms there is breakout above and within the Woodford.

Clay WaterIrreducible

Water GasClay WaterIrreducible

Water GasClay WaterIrreducible

Water Gas

Fig. 10. Detailed display of the Woodford shale pointing out T2 positions for CBW, BVI and gas. The BVI and CBW distributions vary in character within the shale.

Page 7: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

7

Making full use of the early T2 spectrum is now only an expectation. As we collect more MR log data in source rocks, we will be able to establish how much description can be extracted.

COMPARISION OF LOG AND CORE ANALYSIS

Table 1 is the result of core analysis for the same section of Woodford shale. Of primary interest is to compare the TOC values from laboratory analysis with the ΔLogR and LMM results. It is also instructive to note the minerals present in significant quantity. Appendix II shows elemental log results which further confirm findings.

Table 1. Elemental assay of core from an offset Woodford well in the test area yields TOC fractions, by weight, of just under 10%.

CONCLUSION

We have presented a simple, analytical approach to petrophysical evaluation of TOC in shale gas rocks. The method is based upon existing petrophysical techniques and can be implemented using general software packages. The need for a simple method is primarily in lateral/high angle wells where LWD will be used to collect logging data. Furthermore real-time logs will allow well placement remediation and completion decisions to be made before drilling is complete. The simplicity of the solution method fits well with the availability of real-time data. Applied in the Woodford shale we showed that both the LMM and the ΔLogR methods estimate TOC of about 15% by weight. We argue that the LMM is practical and versatile, and fits comfortably within modern petrophysics. However, the ΔLogR method does indicate maturation from increased resistivity, a useful feature. The Woodford shale example demonstrates that our LMM has enough flexibility to deliver TC compatible with the ΔLogR results and the “ground truth” results

from nearby wells and knowledge of TOC in the Woodford. Going forward, it is reasonable to speculate that early time T2 data will discriminate organic from non-organic carbon and indicate organic maturation.

ACKNOWLEDGEMENTS

The authors wish to thank David Jacobi for providing the elemental analysis log shown in Appendix II and we much appreciate helpful discussions with David Jacobi, Brian J. LeCompte, Fred Mendez, John M. Longo, and Matt Bratovich on evaluation of shale resources. Pamela F. Boschee provided invaluable editorial assistance. We thank Baker Hughes, Inc. for support and permission to publish this work.

REFERENCES

Cardott, Brian J. 2008a. Thermal Maturity of the Woodford Shale in Oklahoma Applied to the Gas-Shale Play. Oklahoma Geological Survey. Cardott, Brian J. 2008b. Overview of Woodford Gas-Shale Play in Oklahoma, 2008 Update, Oklahoma Gas Shales Conference. Oklahoma Geological Survey. Comer, J.B. and Hinch, H.H. 1987. Recognizing and quantifying expulsion of oil from the Woodford Formation and age-equivalent rocks in Oklahoma and Arkansas. AAPG Bulletin, 71: 844-858. Fertl, W. and Chilingar, G. 1988. Total organic carbon content determined from well logs. SPE Formation Evaluation, 3 (2): 407-419. Gladkikh M., Hursan, G. Jacobi, D. LeCompte, B. Longo, J. Mendez F. 2008. Evaluating shale gas reservoirs: the Barnett Shale project. Houston Technology Center Research Document, private communication. Hinch, Henry H., and Derby, James R. 1998. Baker Experimental Test Area BH-1 Well Report, internal Baker Hughes document. Jacobi, D., Gladkikh, M., LeCompte, B., Hursan, G., Mendez, F., Longo, J., Ong, S., Bratovich, M., Patton, G., and Shoemaker, P. 2008. Integrated petrophysical evaluation of shale gas reservoirs. Paper SPE 114925 presented at the CIPC/SPE Gas Technology Symposium 2008 Joint Conference, Calgary, Alberta, Canada, 16 -19 June.

Depth feet Si O2 Al2 O3 K2O Fe2 O3 TOC2627 57 16 4.9 4.6 5.262628 59 17 5 4.8 3.732630 59 17 5 4.2 4.752644 54 13 4.4 6 8.692645 53 14 4.3 6.5 8.622646 53 14 4.2 6.3 8.872647 48 12 3.6 12 9.852648 49 12 3.8 9 9.42

Page 8: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

8

Passey, Q., Creaney, S., Kulla, J., Moretti, F., and Stroud, J. 1990. A practical model for organic richness from porosity and resistivity logs. AAPG Bulletin, 74 (12): 1777-1794. Pemper, R., Sommer, A., Guo, P., Jacobi, D., Longo, J., Bliven, S., Rodriguez, E., Mendez, F., and Han, X. 2006. A new pulsed neutron sonde for derivation of formation lithology and mineralogy. Paper SPE 102770 presented at the SPE Annual Technical Conference and Exhibition, San Antonio, Texas, USA, 24-27 September. Prammer, M.G. et al. 1996. Measurements of clay bound water and total porosity by magnetic resonance logging. The Log Analyst, 37 (6).

ABOUT THE AUTHORS

Dr. Daniel F Coope is a petrophysical consultant. He entered the oil industry in 1978 as a Senior Scientist developing LWD sensors and applications. In 1985 he moved into operations and support, working in Aberdeen, London, and more recently, Abu Dhabi. Dan returned to the United States in 2008, where he lives in Annapolis. Dr. Coope holds a PhD in nuclear physics from University of Illinois. He has been a member of the SPWLA since 1980 where he served as a Director at Large and Distinguished Speaker, and the SPE since 1982 where he served as local Chairman of the Houston Drilling Study group. Dan has over 40 publications and 5 patents on petrophysics, applied medical and instrument physics, and nuclear physics. His current interests are source rock evaluation, applications of magnetic resonance, and teaching petrophysics.

Terrence (Terry) Quinn holds a B.S. degree in Chemistry and Physics from Tulane University. Currently Geoscience Applications Engineering Manager for INTEQ’s Answers While Drilling Delivery team working closely with LWD measurements and applications, he began his career by joining Dresser Atlas in 1976 and has worked in operations, log analysis and petrophysics, and marketing. Terry has been very active in the SPWLA as President, VP Technology, as well as Regional Director among many other duties. Elton Frost is Manager for Software Strategy in the INTEQ Answers While Drilling Delivery team in Houston. He joined Dresser Atlas in 1977 and has been involved in formation evaluation since that time. He holds a B.A. and M.A. in Mathematics from Southwest Texas State University.

Michael Manning is Geoscientist Advisor with Baker Hughes INTEQ’s Answers While Drilling Delivery team in Houston. He has previously served on the Technology Committee of SPWLA and chaired the 1990 SPWLA Special Conference on Borehole Imaging. He received his M.S. in Physics from the Pennsylvania State University.

APPENDIX I

In our example, we measure MR (MPHS), bulk density (RHOB) and neutron porosity (NPHI). (Acoustic porosity could also be used, but not in this case). The simplest form of our limited model uses seven unknowns; thus we will relate three independent measurements to seven unknowns. Clearly we need other knowledge and some assumptions. We also choose to employ a minimization approach to deal with measurement uncertainty. We measure:

Unknowns are:

We can safely assume values for water and carbon density. We can also impose mass balance:

We have also implicitly assumed

.

ϕmatrixN is determined from the MPHS vs. NPHI

crossplot (RHOB vs. NPHI could be used). But there is no justification that we can combine carbon and shale neutron porosity response in a single value. The appropriateness of this assumption will be tested over time. However, the error is probably small since fcarbon « fshale in most cases. Consider these typical numbers:

31.NfNf

34N)f(f then

40f and 15,N

38,N .7,0f 0.2,f

shaleshalecarboncarbon

matrixshalecarbon

shalecarbon

matrixshalecarbon

=+

=×+

==

===

ϕϕ

ϕ

ϕ

ϕ

matrix solid 100%in responseneutron is N and fraction, is f porosity, is φ density, is ρ

N)ff(φNPHI

φρfρfρRφMPHS

matrix

matrixshalecarbon

watercarboncarbonshaleshale

ϕ

ϕ

where

HOB

×++=

×+×+×==

ϕmatrixcarbonshalewatercarbonshale N ,f ,f ,ρ ,ρ ,ρ φ,

carbonshale f f φ1 ++=

ϕϕϕshaleshalecarboncarbonmatrixshalecarbon NfNfN)ff( +≅×+

Page 9: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

9

These assumptions and observations reduce the problem to four (4) relations to solve for four (4) unknowns:

.N ,f ,ρ φ, matrixcarbonshaleϕ

Our method is improved by considering the “soft” MR parameters, CBW and BVI. These parameters allow us to over-determine the set of equations and unknowns so as to improve our solution. Finally, these independent measurements do not guarantee that all three are sensitive to the critical unknown, the fraction of carbon present, fcarbon. To address this issue we assign uncertainties to each measurement and solve these equations using error minimization. This does not necessarily lead to a unique solution, but certainly the most likely solution results.

Page 10: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

10

APPENDIX II

Appendix II. Elemental analyses from pulsed neutron-induced inelastic and prompt capture gamma ray spectra are combined with natural spectral gamma ray logs to yield the mineral analysis on the far right and lithology evaluations in the second and third tracks from the right. The Woodford shale has ~700 total Gamma Ray API units, ~70 ppm Uranium from the gamma ray, elemental log-derived organic carbon (black shading) from 9 to 14%, and significant pyrite concentration (red shading), especially in the lower-most part of the Woodford.

Page 11: 2009-23168.pdf

SPWLA 50th Annual Logging Symposium, June 21-24, 2009

11

Fig. 9. Log from well BETA N5 including T2 distribution. The LWD-MR tool (MagTrakTM) is unique in having essentially no gradient. The benefit in this application is that gas should not be expected to occur in the fast decay times, since there is essentially no diffusion.