A Data-Driven Integration of Pore Geometry and Aspect Ratio into Rock Physics Models for Porosity Partitioning and Elastic Property Estimation in Carbonate Reservoirs

Abstract Carbonate reservoirs exhibit complex, heterogeneous pore systems, with pore shapes and sizes spanning nanoscale to meter scale due to primary and secondary porosity, varied fabrics, and textures. Conventional rock physics models, effective for structurally consistent clastic reservoirs, rely on generalized pore geometry assumptions that often fail to capture carbonate variability and therefore require calibration. A data-driven approach enhances Differential Effective Medium (DEM) models for carbonate reservoir characterization by integrating empirical pore system data. A robust real-world dataset comprising over 1,750 thin sections from five wells—one of the largest reported for carbonate reservoirs—provided the foundation for image analysis to quantify pore-type volume fractions and aspect ratios, directly informing the model. Analysis revealed that reference pores (interparticle porosity) dominate, with crack-like pores comprising less than 0.1% of the total pore population, challenging typical model assumptions. The pore geometry data were incorporated into rock physics models through optimization techniques, establishing a valuable aspect ratio database for carbonates. Unlike conventional approaches that rely on predefined pore categories (reference, crack, stiff) and treat aspect ratios as fitting parameters or constant values assigned to each pore type within a target layer, aspect ratios were derived directly from measured data, enabling precise porosity partitioning. These insights supported evaluation of elastic property variability, quantification of pore geometry’s impact on compressional and shear velocity predictions, and assessment of pore type distribution across the reservoir. Optimizations based on compressional velocity alone, shear velocity alone, and both velocities jointly yielded R2 values exceeding 90%, indicating high predictive accuracy. The approach also revealed scale-dependent fracturing effects, refining reservoir characterization. This method addresses conventional model limitations, improving porosity partitioning and elastic property estimation in complex carbonate reservoirs while providing a framework applicable to future studies.