Mapping permeability in low‐resolution micro‐CT images: A multiscale statistical approach

We investigate the possibility of predicting permeability in low-resolution X-ray microcomputed tomography (µCT). Lower-resolution whole core images give greater sample coverage and are therefore more representative of heterogeneous systems; however, the lower resolution causes connecting pore throats to be represented by intermediate gray scale values and limits information on pore system geometry, rendering such images inadequate for direct permeability simulation. We present an imaging and computation workflow aimed at predicting absolute permeability for sample volumes that are too large to allow direct computation. The workflow involves computing permeability from high-resolution µCT images, along with a series of rock characteristics (notably open pore fraction, pore size, and formation factor) from spatially registered low-resolution images. Multiple linear regression models correlating permeability to rock characteristics provide a means of predicting and mapping permeability variations in larger scale low-resolution images. Results show excellent agreement between permeability predictions made from 16 and 64 µm/voxel images of 25 mm diameter 80 mm tall core samples of heterogeneous sandstone for which 5 µm/voxel resolution is required to compute permeability directly. The statistical model used at the lowest resolution of 64 µm/voxel (similar to typical whole core image resolutions) includes open pore fraction and formation factor as predictor characteristics. Although binarized images at this resolution do not completely capture the pore system, we infer that these characteristics implicitly contain information about the critical fluid flow pathways. Three-dimensional permeability mapping in larger-scale lower resolution images by means of statistical predictions provides input data for subsequent permeability upscaling and the computation of effective permeability at the core scale.