SPWLA 45th Annual Logging Symposium, June 6-9, 2004 AN ADVANCED EVALUATION METHOD FOR LAMINATED SHALY SANDS INCLUDING UNCERTAINTY AND SENSITIVITY Johan van Popta and Paul Hofstra (Shell EP Technology) and Steven van Houwelingen (Data Comprise) ABSTRACT Laminated sand/shale reservoirs can be found in, for example, fluviatile, deltaic and, most importantly, deep-water turbidite reservoirs. A petrophysical evaluation of such a reservoir has to meet several requirements, such as: timely detection of pay intervals, accurate evaluation of hydrocarbon volumes and production potential, and appropriate estimation of uncertainty This paper describes a petrophysical evaluation program that has been developed to meet all those requirements. It uses a petrophysical model characterized by four volumes. For the invaded zone these are volume of wet shale, volume of sand grains, volume of mudfiltrate and volume of hydrocarbons. For the virgin zone the last two are replaced by the volume of formation water and of hydrocarbons, ‘These volumes are linked to the properties of a laminated sand/shale model using the Thomas-Stieber method, The resistivities of the sand and shale laminations within the tool vertical resolution can be combined, assuming either an anisotropic formation with parallel layering or an isotropic formation wherein the layering is absent or distorted, Basic nuclear and resistivity log data are sufficient input to apply the program, Evaluation results of borehole image tools and advanced NMR and multi-component or triaxial induction logging data can be combined with basie logs to achieve a consistent evaluation and a reduction in uncertainty. The combination of all input logs is done through a flexible quality-weighted and constrained inversion method. The output consists of the volume fraction of sand laminations and the sand lamination porosity, dispersed clay volume and hydrocarbon saturation plus some auxiliary results From user-specified uncertainties in input logs and. valuation parameters the program evaluates with a Monte Carlo method a cumulative probability distribution of the total hydrocarbon volume over an. interval. In addition, the sensitivity of the evaluated total hydrocarbon volume to individual uncertainties in input log data and evaluation parameters can be estimated and ranked. This enables the user to focus the total evaluation effort on the inputs that are of highest importance. The paper also presents examples of application to field data in order to illustrate the capabilities of the ‘method in producing accurate results including, quantitative estimates of the range in results. INTRODUCTION For a long time the evaluation of laminated shaly sand reservoirs has posed difficulties. The main reason is the strong and non-linear suppression of horizontal resistivities by conductive shale laminations. Asa result the presence of hydrocarbons is sometimes difficult to recognize and a proper evaluation of the hydrocarbon saturations in the sand laminations requires use of a suitable model. Resistivity anisotropy in layered systems has been recognized and described by for example Moran and Gianzero (1979). Petrophysical models for laminated sand-shale reservoirs have been proposed by for example Thomas and Stieber (1975), Haley (1979), ‘Van den Berg (1996), Klein (1997), Mollison (1999) and Schén (1999). Besides using a suitable petrophysical model there is need for a method that can use as many log data and core data as are available in a consistent manner to provide confidence in the results and reduce uncertainty. A general method is described by Quirein (1986) and a method more specific for laminated sand shale reservoirs is described by Mezzatesta (2002). This is all the more important with the advent of advanced logging tools, such as NMR tools or multi- component induction logging tools. RRI Further, there is need for a method that is able to quantify the uncertainty range in the results as caused by uncertainty in input log data and in evaluation parameters used in the petrophysical model. In order to reduce uncertainty in the results the petrophysicist needs a tool to analyse the impact of uncertainty in ceach log measurement and each evaluation parameter individually. A comparison of different methods for uncertainty evaluation is provided by Verga (2002).SPWLA 45th Annual Logging Symposium, June 6-9, 2004 This paper reports on a program that was developed to ‘meet all three requirements. METHODOLOGY The program uses a formation model characterized by the following volume variables: wet shale, sand grain, and invaded zone fluids mudfiltrate and hydrocarbons. In the virgin zone the fluids are formation water and hydrocarbons. Figure 1 displays a schematic model of the volumes. The formalism employed is as was used by Van den Berg (1996) and expressed by Mollison. (1999), These volumes are determined by iterative forward modeling of the input logs till a satisfactory match is obtained. Calculation of the response of density and neutron logs takes invasion effects into account. The calculation of resistivities incorporates the parallel and series addition of unresolved shale and sand laminations. As an alternative to this anisotropic resistivity model, the user ean also choose an isotropic resistivity model. Such a model can be applicable to situations where the laminations are distorted by slumping or bioturbation. It should be noted that such a distorted layering cannot be recognized from a Thomas-Stieber plot. The program can include as input logs horizontal and vertical resistivities, as derived from data acquired by three component induction logging tools or from LWD propagation resistivities in high angle wells, and NMR clay bound water and total porosity curves together with conventional logs in an integrated evaluation. The user has flexibility to include or exclude logs, thereby evaluating the impact of such logs on the evaluation result, and (0 weigh each log by assigning an accuracy to its response. The evaluated volumes are used to derive the sand lamination total porosity and the dispersed clay volume, using the Thomas-Stieber approach, and the sand lamination saturation, Conductivity effects of dispersed clay in the sand laminations are ineluded using the approach of Waxman and Smits (1968). The minimisation method is a constrained and weighted least squares method, where the weight is inversely proportional to a user supplied log error estimate. The constraints arc related to allowed areas in the Thomas-Stieber shale distribution plot and to invasion and also the user can supply a maximum for the sand lamination resistivity The error function that is minimized, called incoherence, has the form: io -r,) J where: input, input log j rec reconstructed log j estimate of standard deviation for ogi ‘The incoherence value and the reconstructed logs at the last iteration are program outputs which help the user to assess the quality of the inversion, The quantification of uncertainty is in the equivalent hydrocarbon column (EHC) which is: LOPS where y, Volume of laminar shale $43 porosity ofthe sand lamination Shy hydrocarbon saturation of the sand lamination, ‘The sum is taken over a user selected interval and the EHC over this interval is a measure of hydrocarbons. in place. The uncertainty range is evaluated in & Monte Carlo fashion by adding random noise to input parameters and input logs according to user supplied standard deviations and redoing the evaluation, Experience learns that about 100 - 200 evaluations are adequate to achieve a representative distribution of the EHC values. ‘The sensitivity of the EHC to cach evaluation parameter and log can be evaluated by adding a representative noise value in the log or parameter and reevaluating the interval. This enables the user to identify the uncertainties that have the biggest contribution to the EHC spread. The user can assign correlations or dependencies between input parameters or between input parameters and logs by providing a correlation matrix. An example is the correlation between the resistivity log and the shale resistivity read from the log over representative shales. Any calibration noise on the resistivity log will apply to the shale resistivity as well. By imposing thiscorrelation the contribution of noise on the resistivity log to the spread in EHC will be reduced. EXAMPLE 1 ‘The well used in the first example has penetrated a deep-water oilbearing turbidite reservoir containing massive sands and an extended section of low resistivity pay with a shale content of around 50% and resistivities that are clearly below those of the massive sands, but above the shale baseline and well above the wet sand resistivity of around 0.2 Ohmm. The well ‘was drilled with oil-based mud at a fairly low deviation of 20 degrees. A display of basic log data and evaluation results is shown in Figure 2 From regional geologic knowledge and reservoir production performance it is known that this interval is very likely laminated with large lateral extent and connectivity of the laminations. The Thomas-Stieber plot displayed in Fig, 3 shows clearly a laminated distribution with little dispersed clay. This implies that the porosity in the laminated sands is very similar to the porosity of the clean sands. Based on all this information the evaluation model chosen was a anisotropic resistivity mode! with a shale distribution as derived from the Thomas-Stieber method. The resulls obtained with this evaluation model are displayed in Fig. 2. These results show that the porosities and oil saturations of the sand laminations are very similar to those of the massive sands. This similarity suggests that the laminated interval has similarity as well in sand permeability, which ideally is validated by, for example, grain size distribution data from side wall samples or an NMR log which likely will need calibration to derive the sand lamination permeability. ‘Two hundred Monte Carlo simulations for the equivalent hydrocarbon column (EHC) over the laminated part (18,258 - 18,376) produce an average EHC of 17.4 ft and a standard deviation of 1.1 fl In this case the uncertainty in the evaluation results is reasonably small for a laminated interval. The reason is that the laminated interval in this well has distinetly higher resistivities than the shale resistivity, which makes the evaluation fairly robust. Another factor is the considerable knowledge on the geology and the production performance of this type of reservoir. It is to be realized that this uncertainty estimate is only valid within the model of an anisotropic resistivity SPWLA 45th Annual Logging Symposium, June 6-9, 2004 The results of the sensitivity analysis of the evaluated EHC are shown in Fig. 4. Table 1 provides the uncertainty input data for the logs and the most important evaluation parameters. These results show that in this case the shale resistivity and density are the most important sources of uncertainty. Its to be realized that the choice of representative shale parameters is not trivial as the laminated shales are not resolved by the logs. Selection of a shale, thick enough to be resolved by the logs, that is representative for the laminated shales requires care and is best done in close cooperation with a geologist EXAMPLE 2 The second well, drilled with oil-based mud at a deviation around 35 degrees, has penetrated a reservoir with low resistivity pay in the upper part. In this type of reservoir laminated intervals are quite common, but there is risk of bioturbation which distorts the laminated layering and consequently the productivity. Figure 5 displays a depth plot of log data and evaluation results. In this well a multi-component induction log was run, the 3DEX, to obtain horizontal and vertical resistivities. A Thomas-Stieber plot for the interval 350 - 450 ft is displayed in Fig. 6. Over {interval 400 - 450 ft this plot shows mainly high N/G sands with low dispersed clay, Over the upper part, interval 350 - 400 f, the data represent a laminated distribution in which the sand laminations have considerable dispersed clay content. The main interval of interest for this evaluation is the upper interval and the questions are how much hydrocarbons are present in this interval and what could be its productivity Three evaluations are made to assess the uncertainty: 1. The interval was evaluated using the density, neutron and all three resistivity logs, Rt, horizontal and vertical resistivity from the 3DEX. The assumption of an anisotropic reservoir was made, with the amount of anisoiropy being determined by the 3DEX horizontal and vertical resistivities. The results of this evaluation are shown in columns 5 till 7 and the curve hyd voll, which equals (I- Vlam)*por_sand*sh_sand, in column 8 2, Am evaluation was made using for resistivity only the Rt log and using an anisotropic layered reservoir model. The resulting hyd vol2 curve is displayed in column 8. This result represents an optimistic case in which the layering has good parallel bedding without distortion by bioturbation. RRISPWLA 45th Annual Logging Symposium, June 6-9, 2004 3. An evaluation was made as under 2 but inthis ease using an isotropic reservoir model. The resulting hyd vol3 curve is again displayed in column 8. This represents a pessimistic case in which the layering is completely distorted by bioturbation or in which the sands have a low resistivity, in this ease taken equal to the resistivity of the Rt curve. For evaluation #1 also a distribution of EHC values. was produced. This distribution together with single evaluation results for the evaluations #2 and #5 are presented in Fig. 7. The most important conclusion from Fig. 7 is that the evaluation #1 result falls outside the distribution produced using the horizontal and vertical resistivity, Use of these resistivities in the evaluation rules out the pessimistic model of a isotropic reservoir in which the sands have very low hydrocarbon saturations. The second conclusion is that at these low saturations the uncertainty range becomes relatively large. The third conclusion is that the productivity of this interval is expected to be low, because of the dispersed clay and the reduced sand lamination porosities as seen in the evaluation results and the Thomas-Stieber plot CONCLUSIONS The examples illustrate that this evaluation program for both laminated and anisotropic sand-shale reservoirs and isotropic sand-shale reservoirs has @ wide functionality, including flexibility in using a variety of input logs. tis able to include uncertainty in input logs and petrophysical evaluation parameters and estimate a representative range of results. Also, it provides the user with a tool to rank uncertainty in input data according to highest impact on the valuation result. These are attractive features in a time when quantifying and reducing uncertainty receives a high amount of attention. ACKNOWLEDGEMENTS We thank the Shell operating units that acquired the field data for permission to publish the data, Also the contribution of many Shell colleagues through insightful discussions is gratefully acknowledged REFERENCES Berg, F.G. van den, Looyestijn, W.J., Sandor, RK, SANDWICH: log evaluation in laminated shaly sands, SPWLA 37° Annual Logging Symposium ‘Transactions, 1996, Paper BB. Haley, R.A., A synergetic log and core analysis program using a laminated shale - dispersed elay sandstone model, SPWLA 20 Annual Logging Symposium Transactions, 1979, Paper F Klein, J.D., Martin, P.R., and Allan, D-F., 1997, The Petrophysics of electrically anisotropic reservoirs, The Log Analyst, May-June, pp.25-36. Mezzatesta, A.G., Rodriguez, E-F., Mollison, RA. and Frost, E., Laminated shaly sand reservoirs - An. interpretation model incorporating new measurements, , SPWLA 43" Annual Logging Symposium Transactions, 2002, Paper TT Mollison, R.A., Sebi, J., Fanini, ., Kriegshiuser, B., Meyer, H. and Gupta, P., 1999, A model for hydrocarbon saturation estimation from an orthogonal tensor relationship in thinly laminated anisotropic reservoirs, SPWLA 40” Annual Logging Symposium ‘Transactions, Paper 00. Moran, J.H., and Gianzero, S., 1979, Effects of formation anisotropy on resistivity logging measurements, Geophysics 44 (luly 1979): pp. 1266- 1286. Quire, J, Kimminau, S,, LaVigne, J, Singer, 1 and Wendel, FA coherent framework for developing and applying multiple formation evaluation models, SPWLA 27" Annual Logging Symposium Transactions, 1986, Paper DD. Schén, J.H., Mollison, R.A. and Georgi, D.T., 1999, Macroscopic electrical anisotropy of laminated reservoirs: A tensor resistivity saturation model, SPE paper $6509, SPE Annual Technical Conference ‘Thomas, E.C. and Stieber, $.J., The distribution of shale in sandstones and its effect upon porosity, SPWLA 16" Annual Logging Symposium ‘Transactions, 1975, Paper T. ‘Verga, F., Viberti, D., Gonfalini, M., Uncertainty in ‘well logging: Analytical or numerical approach? SPWLA 43" Annual Logging Symposium ‘Transactions, Paper C. Waxman, M.H., and Smits, L.J.M., 1968, Electrical conductivities in oil bearing shaly sands, SPW Journal 8, No.2, 107-122ABOUT THE AUTHORS, Johan van Popta received a Mse. degree in Physics in 1977. In 1979 he joined Shell as a research petrophysicist. During his employment in Shell he ‘worked on a variety of research projects, such as quantitative seismic interpretation, resistivity and borchole gravimeter logging and laminated sand evaluation methods as well as in operational petrophysics and field studies. Currently he works in a reservoir surveillance group. Paul Hofstra has a PhD in Theoretical Nuclear Physies from the Vrije Universiteit in Amsterdam, Since 1980 he worked for Shell, mainly in Research, From 1984 until 1990 he spent time as operational and SPWLA 45th Annual Logging Symposium, June 6-9, 2004 studies petrophysicist in Brunei and Germany. His main interest is in well log interpretation, subsidence and inversion, Steven van Houwelingen (1969) has received his Mc, in Electrical Engineering at Delft University of Technology in 1993. Hereafter he followed a two-year post-doe program at the Statistics Stochasties and Operations Research group of the math department of | the same university. During and after this program he contributed as a contractor to Resistivity Tool ‘modeling and software development in Shell EP Technology. Parameter igma Unit ‘Sensitivity in EHC (%) horizontal shale 2.00E-01 fraction (relative “69 resistivity change) [density shale 4.00E-02 gramicc 53 [density log 2.00E-02 ‘gramicc 49 neutron log 3.00E+00 percent 19 density grain 2.00E-02 ‘gramice 18 [deviation angle 5.00E+00 degrees “15 [deep resistivity 1.00E-01 fraction (relative 14 change) ‘Table 1 Uncertainty in the input logs and the most important parameters in the evaluation and the resulting sensitivity in the evaluated EHC. (example 1) RRISPWLA 45th Annual Logging Symposium, June 6-9, 2004 <— ben > shale laminae, volume Vise, sand laminae, volume (1 — Vim) bie be sand grain dispersed clay hydro- formation volume clay grain water carbon water volume volume volume volume ‘Figure 1 ‘The formation model for the virgin zone that is used. Within the Thomas-Stieber method this ‘model ean be described by four volumes: Volume of formation water, volume of hydrocarbons, volume of wet shale, volume of sand grains. The assumption is that laminated and dispersed shale have the same porosity. Within the Thomas-Stieber method the sand lamination ean contain either dispersed or structural sha.SPWLA 45th Annual Logging Symposium, June 6-9, 2004 Tneeh Density Ten eh! [pore vol [he vel a 1 2.85 2.65 a alo alo 2 Gonme Ray | “pertH: | Neutron Rt tot ent | ed por | oa eat j@_ API 1e0| 2-/see- [4a PU oO]. S OHM Sala ala ala 2 ne { f \ t H } ‘ | xan % 3 Figure 2 Display of log data and evaluation results for well 1. The incoherence plotted in the left track is mostly close to | and always less than 2 indicating a good quality reconstruction of the logs. Track 5 displays the total and laminated shale content, the difference between them is mostly dispersed clay and sometimes structural clay. Track 6 displays the sand lamination porosity and the sand lamination pore volume, (1-Vlam)*por_sand. Track 7 displays the sand lamination oil saturation and the hydrocarbon volume, (1-Vlam)*por_sand*sh_sand. RRISPWLA 45th Annual Logging Symposium, June 6-9, 2004 high NIG, | | clean sands Varying NIG, mainly é laminated | / 1» [sbale, clean andl | high porosity sand laminations a 8 Densty (g/cc) Lower NiG, laminated and dispersed shale, sand laminations have porosity reduced by dispersed clay Figure 3 Thomas-Stieber plot of density and neutron log for example 1. Note the predominantly laminated sequence. Parameter orlog Sensitivity in EHC (4) Figure 4 Results ofa sensitivity evaluation for the EHC for example 1. Only the most important sources of uncertainty in the EHC are shown.SPWLA 45th Annual Logging Symposium, June 6-9, 2004 iz. =] iz _Ouvn 28 226} Cotiper Neutron | ver Res | vlom_[eore vel mga vole ls IN 26] Isa PU efx Ou 29) = e060 anna Rou | ‘vere: | Daneity | hor Rea | Usnote | eond per | ed sat | nya vols XX350 ) Xx400 XX450 Figure $ RRI Display of log data and evaluation results for well 2. Track 4 displays the resistivity logs, the Rt curve from an array induction tool and the horizontal and vertical resistivity from a multi- component induction logging tool, the 3DEX. The interval of interest for this evaluation is 350 - 400. Track 8 displays the hydrocarbon volume, (1-Vlam)*por_sand*sh_sand, curves resulting, from three evaluations,SPWLA 45th Annual Logging Symposium, June 6-9, 2004 ‘Neutron free) a ‘ read high NIG, Ay XX400-XX450 i : clean sands, low | J | dispersed clay AN | > 23 | 23 a. 1 hha Laminated interval, XX350 a | —XX400; sand laminations have porosity reduced by dispersed clay uo 0 03 oF Os 8 7 Figure 6 ‘Thomas-Stieber plot of density and neutron log for example 2, Note the presence of dispersed clay in interval 350 - 400SPWLA 45th Annual Logging Symposium, June 6-9, 2004 100 . Evaluation Evaluation Lf #3 result #2 result 80 ++ + 2 2 0 £0 i ° 20 20 of a 00 os 10 15 20 2s EHC ( Figue 7 EHC results for interval 350 400 fas evaluated by three methods including the distbution obtained with method 1 RRI