Geostatistical inversion integrated with rock physics modeling to amplify thin reservoirs

Abstract

Conventional reservoir characterization methods struggle to adequately characterize the widely spread, prolific, thin, and heterogeneous gas sand layers of the Lower Goru Formation (LGF) in the Lower Indus Basin (LIB). This study unveils the potential of advanced techniques to decipher the hidden potential within the LGF challenging reservoir intervals, by the multi pronged approach. The petrophysical analysis identified two potential zones (Sand-A and Sand-B) within the B-Interval with good porosities exceeding 10%. Rock physics modelling (RPM) developed an extended cemented sandstone model to estimate the missing shear sonic log (Vs) and optimize the density log. The improved modeled elastic properties logs (P impedance, Vp/Vs ratio, Lambda-rho (λρ), Mu-rho (μρ)) crossplots help to highlight reservoir properties and differentiate pay and non-pay facies robustly. Low impedance (5000–9000 (g/cc*ms)), Vp/Vs ratio range (1.4-1.6), and Poisson ratio (0.1-0.25) along with less than or equal to λρ 28 (GPa*g/cc), and greater than or equal to high μρ 60 (GPa*g/cc) designated gas bearing intervals in heterogeneous tight sands. Statistical metrics, such as good prediction quality, a high correlation coefficient, and a lower normalized root mean square error for modeled logs P-sonic (Vp), Vs, and density (ρ) also validated the accuracy of the chosen model. Pre-stack simultaneous seismic inversion (PSSI) highlighted thin gas-bearing intervals through low P-impedance (Zp), S-impedance (Zs), and Vp/Vs ratio properties volumes. PSSI proved insufficient in certain localities to distinguish thin gas-bearing sands from non-gas-bearing zones despite exhibiting low impedance values within the B-Interval sand layers attributed to the "thin bed tuning effect" as gas-bearing sands have an average thickness of 14-17 meters. The maximum estimated tuning thickness from a synthetic wedge model is 55-60 meters for the employed seismic data. Bayesian stochastic seismic inversion (BSSI) surpassed conventional characterization techniques by leveraging low and high frequencies extracted from optimized well data. This enabled the inversion of a 3D seismic dataset into a finely spaced stratigraphic grid (0.1 ms intervals), unlocking high-resolution elastic attributes that highlight a detailed picture of the reservoir properties. The RPM improved elastic properties in combination with petrophysical properties using probability density functions (PDF’s), allowing high-resolution elastic properties transformed into reservoir features (geological facies). Consequently, applying RPM logs along with BSSI discovered thin heterogeneous gas sands probabilities precisely. This study further presents a comprehensive approach to reservoir characterization by combining the power of machine learning (ML) with continuous wavelet transform (CWT) applied to seismic and well data. CWT sharpened the resolution by xxiv | P a g e highlighting subtle reservoir features hidden within the seismic signals. The CWT’s enhanced features have fed into the best-performing ML algorithms, Extra Tree Regressor (ETR), and Decision Tree Classifier (DTC), trained on a rigorously calibrated dataset (65% training, 35% random test data, and blind well validation). ETR effectively predicted petrophysical and elastic characteristics with high accuracy. Similarly, DTC outclassed the other algorithms with the highest scores for precision (89%), recall (92.3%), and F1 (90.6%) for gas-sand facies at blind well locations. It developed probability maps revealing the potential of thin heterogeneous gas-sands. This integrated approach provides a powerful tool for unlocking the challenges of thin gas sands in complex basins. It holds the potential for optimizing exploration and production by accurately targeting gas-bearing zones and minimizing drilling risks. It helped advance understanding of LGF by gaining insights into its depositional environment and reservoir heterogeneity. This study paves the way for more efficient and successful exploration and development of similar thin heterogeneous gas reservoirs worldwide.