2405 rue de la Terrasse
Defined as soil or rock that remains at or below 0 °C for a long period of time, permafrost (i.e. perennially frozen ground) covers extensive areas in Arctic regions. It occurs in practically all types of geological surface material such as solid, fractured and weathered bedrock, gravel, sand, silt, clay or peat. Permafrost contains ice in various forms and amounts. Structural and thermal design considerations when building in the Arctic require precise knowledge of the thermal and geotechnical properties of permafrost. Property values are also necessary as input for the parameterization of heat transfert models and thaw settlement prediction. Previous studies (Calmels and Allard, 2008 and Calmels, 2010) showed great potential in using computed tomography for classification and volume measurements of permafrost components, i.e. sediment (solid), ice and gas (void) contents. The technology also provides visualization of the structural organization of permafrost (cryostructure) and to some extent, depending on system resolution, of more intimate soil particles-ice organization (cryotexture). A new approach to measure permafrost thermal conductivity is being developed combining proven thermal conductivity models (Schwerdtfeger, 1963; Farouki, 1981; Côté and Konrad, 2005) and computed tomography analyses. The aims of my study are to (1) present the application of an innovative and non-destructive approach using CT-scan to estimate thermal conductivity of undisturbed permafrost samples and (2) validate the results computed from CT-scan image analysis with experimental thermal conductivity tests. I use a three-step model that takes into account the soil type, the porosity of ground ice and the cryostructure (forms of ground ice) in the samples to assess the potential of the proposed method. To do so 20 permafrost samples with different textures and cryostructure (French and Shur, 2010) (plus one massive, bubbly ice core), ranging from homogeneous fine-grained soils with stratified ice lenses to coarse-grained diamictons well-bonded with pore ice, were extracted from various sedimentary environments (glacial, alluvial, marine, organic, etc.) in the Nunavik and Nunavut regions. The core samples were scanned using a Siemens Somatom 64TM scanner at the Institut National de la Recherche Scientifique (INRS) in Québec city. The samples were scanned over their entire length with slice thickness of 0.4 mm. According to the core diameter (100 mm), pixel resolution of 0.1 x 0.1 mm was obtained. By selecting a range of TI values corresponding to each of the soil components (sediments, ice, and gas) (Clavano et al., 2011), voxel classification and quantification of the sample components were achieved using ORS Visual © software of Object and Research Systems, therefore providing volumetric contents of the frozen sediments and the bubbly ice (xsoil, xbi) in the permafrost samples. The thermal conductivity tests are conducted at the Laboratoire de géotechnique de l’Université Laval. Following the same experimental setup as Côté and Konrad (2005), the tests are conducted inside a cell surrounded by an insulated box at a constant temperature of about -8 °C. Temperature boundary conditions at the top (-4 °C) and bottom (-12 °C) of the cores were maintained with two independent heat exchangers creating a vertical heat flow through the sample (Coté and Konrad, 2005). In order to assure a flawless contact between the sample and the copper thermocouple, a weight is placed on the top of the experimental setup. Each core is tested over a minimal period of 24 hours. The CT-scan used in this study provides voxel resolution larger than the porosity of fine-grained sediments such as silt and clay, yielding an underestimation of pore ice and air content (Eqn. 6) which is affecting the prediction of effective thermal conductivity. Nevertheless, the comparative results between CT-scan derived conductivities and thermal conductivity cell results show a great potential for the method. As a preliminary study, 7 out of 9 tested samples showed a margin of error of less than 20%. More validation work is also planned by conventional laboratory work such as, particle size analyses, water content determinations, methylene blue test and mineralogy. These measured physical properties shall further improve our understanding of how composition and cryostructure affect thermal conductivity from CT-scan results.