Challenges and uncertainty in the seismic reservoir characterization of Bone Spring and Wolfcamp formations in the Delaware Basin using rock physics Ritesh Kumar Sharma †* , Satinder Chopra † , James Keay ‡ and Larry R. Lines + † TGS, Calgary; ‡ TGS, Houston; + University of Calgary, Canada Summary Amongst other things, rock physics analysis is usually carried out for estimating the volume of clay, water saturation and porosity using seismic data. Though these rock-physics parameters are easy to compute for conventional plays, there are a lot of uncertainties in their estimation for unconventional plays, especially where multizones need to be characterized simultaneously. We discuss them with reference to a dataset from the Delaware basin where the Bone Spring, Wolfcamp, Barnett and the Mississippian formations are the prospective zones. We elaborate on the challenges and the uncertainties in the characterization of these multi-zones, and how we overcome them. Our conclusion is that any deterministic approach (single rock-physics model) for characterization of the target formations of interest may not be appropriate and we build the case for adopting a robust statistical approach, comprising a graphical crossplot method and employing Bayesian classification. While the former makes use of neutron and density porosity data for defining the different lithofacies, the latter yields the uncertainty associated with the individual lithofacies. In this whole exercise, we begin with well-log data and define different lithofacies based on the graphical crossplot method. Thereafter, we correlate these facies with the interpreted mud-log data available for one well. Having gained the confidence in defining the different lithofacies, we then determine the lithofacies and their probabilities using the seismic impedance inversion attributes. The resultant facies seem convincing and correlate well with facies information derived from mud-log data interpretation. Challenges and uncertainty in the characterization of shale formations using rock-physics analysis Rock-physics analysis consists of two parts namely modeling and inversion. As the names suggest, attempts are first made to model the elastic response using mineral fractions, water saturations and porosity. Thereafter, rock-physics properties mentioned above are extracted using elastic properties computed using seismic impedance inversion. As per rock physics, the elastic modulus (M) of a rock can be expressed as follows 1  =∑ (1−∅)    + ∅     . (1) where Mi are the i th mineral moduli and Vi are the i th mineral volume fraction. As can be gauged from the equation above, parameters such as mineral volume fraction, water saturation and porosity play an important role in the rock-physics analysis. While these parameters are relatively easy to estimate for conventional reservoirs, there are a lot of uncertainties in their estimations for unconventional reservoirs. Some of the challenges are discussed below. Uncertainty in the estimation of volume of shale from well-log data Based on the fact that shale is usually more radioactive than sandstones and carbonates, the gamma-ray log curves are used to distinguish shale formations (with higher values) from others. Not only that, gamma-ray logs can also be used to determine the volume of shale present in a formation. Of course, there are other ways of computing the volume of shale from different well-log curves, but gamma-ray logs happen to be one of the methods, where first gamma-ray index is computed and is then transformed into volume of shale using linear or nonlinear empirical relationship. The gamma ray index is defined as I GR=(GRlog- GRmin)/(GRmax-GRmin); I GR represents gamma-ray index, GRlog represents the gamma-ray reading at any depth, GRmin represents the minimum gamma-ray value which would correspond to clean sandstone, GRmax represents the maximum gamma-ray value which would correspond to shale. Thus, one needs at least one or more points on a clean sand, and similarly some points on a real shale rock in the shale interval under investigation. In the absence of such values, the computation could fall apart. For bringing in accuracy in such calculations, a linear correction through the use of a scalar multiplication has been suggested but results in an overestimation of Vsh. Empirical nonlinear corrections have been suggested by Larionov (1969), one for Tertiary or younger rocks, and another one for older rocks (Asquith and Krygowski, 2004). Some other corrections by Stieber (1970) and Clavier (1971) have also been proposed. All these corrections result in improved estimates in certain situations, but inaccuracies still show up in shaly sand formations. However, such empirical corrections have the drawback that they require other independent log curves or core data for calibration. In order to capture the differences among different approaches we implement them on well-log data over a 3D seismic volume from Delaware Basin. In Figure 1, the sonic, density and gamma-ray curves from a well are shown in tracks 1, 2 and 3. The red curves show the input curves as such and the blue curves are their smoothed versions, which were used in the computations. In track 4, the computed volume of shale curve is shown in red, along with the scaled curve in blue and the curve with Stieber correction in black. Notice the large variations in these curves which will introduce discrepancies in the computations they are used in. The volume of shale was computed by a petrophysicist by first subdividing the curves into five basic zones, with the prominent ones being the Bone Spring, Wolfcamp and the Barnett/Mississippian intervals. Next, the minimum and maximum values of gamma-ray log in the respective zones were picked up. Finally, the computations of gamma-ray index were merged into a single composite curve, shown in track 5. This turns out to be different from the other curves shown in track 4. Thus, we see there is uncertainty associated with the determination of volume of shale depending on the type of method adopted. The rule of thumb is to use minimum value of Vsh estimated using above approaches or the one which shows the maximum correlation with available XRD data. 10.1190/segam2019-3214084.1 Page 4928 © 2019 SEG SEG International Exposition and 89th Annual Meeting Downloaded 08/12/19 to 192.160.56.248. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/