2D and 3D imaging resolution trade-offs in quantifying pore throats for prediction of permeability

Although the impact of subsurface geochemical reactions on porosity is relatively well understood, changes in permeability remain difficult to estimate. In this work, pore-network modeling was used to predict permeability based on pore- and pore-throat size distributions determined from analysis of 2D scanning electron microscopy (SEM) images of thin sections and 3D X-ray computed microtomography (CMT) data. The analyzed specimens were a Viking sandstone sample from the Alberta sedimentary basin and an experimental column of reacted Hanford sediments. For the column, a decrease in permeability due to mineral precipitation was estimated, but the permeability estimates were dependent on imaging technique and resolution. X-ray CT imaging has the advantage of reconstructing a 3D pore network while 2D SEM imaging can easily analyze sub-grain and intragranular variations in mineralogy. Pore network models informed by analyses of 2D and 3D images at comparable resolutions produced permeability estimates with relatively good agreement. Large discrepancies in predicted permeabilities resulted from small variations in image resolution. Images with resolutions 0.4 to 4 mu m predicted permeabilities differing by orders of magnitude. While lower-resolution scans can analyze larger specimens, small pore throats may be missed due to resolution limitations, which in turn overestimates permeability in a pore-network model in which pore-to-pore conductances are statistically assigned. Conversely, high-resolution scans are capable of capturing small pore throats, but if they are not actually flow-conducting predicted permeabilities will be below expected values. In addition, permeability is underestimated due to misinterpreting surface-roughness features as small pore throats. Comparison of permeability predictions with expected and measured permeability values showed that the largest discrepancies resulted from the highest resolution images and the best predictions of permeability will result from images between 2 and 4 mu m resolution. To reduce permeability underestimation from analyses of high-resolution images, a resolution threshold between 3 and 15 mu m was found to be effective, but it is not known whether this range is applicable beyond the samples studied here. (C) 2013 Elsevier Ltd. All rights reserved.