An Experimental Study of Thermal Properties of Permafrost Soils

Soil is a multiphase porous media, and its complexity increases when the soil freezes. The high compressive strength of frozen soil has been utilized by geotechnical engineers in construction of frozen earth structures. However, in permafrost settings, the upper layer of the soil is subject to seasonal thawing/freezing and is associated with stability problems. The thickness of this seasonally active layer is dependent on thermal properties of the soil, and the changes in these properties with temperature. Therefore, thermal properties of the soil must be evaluated in frozen ground engineering design. Thermal conductivity is essential for the heat transport in the ground, and reliable data on the thermal conductivity is required. Very often such information is not readily available. The solution is to perform field investigations and laboratory work in order to obtain index parameters. From the obtained soil parameters, the soil thermal conductivity can be estimated from existing empirical equations and analytical models. This process could be inaccurate and time consuming. Hence, direct and efficient methods to directly measure the thermal conductivity are desired. In this thesis, a KD2 Pro Thermal Properties Analyzer has been used to directly measure the thermal properties of frozen and unfrozen core samples from Adventdalen, Svalbard. Laboratory investigations of the soil were performed at the University Centre in Svalbard (UNIS). Index tests include the estimation of water content, bulk density, density of solids, salinity, organic content, degree of saturation, and grain size distribution. Plots of index and thermal properties with depth, were created to investigate dependencies and trends. The tested soils are well-sorted deposits, dominated by fines. The upper part of the depth profile is dominated by high water content, which is a result of accumulation of water due to freezing and thawing processes. The thermal properties have small variations over the profile. However, an increase in thermal conductivity with higher water content in frozen state is observed. The impact of salinity and organic content on the thermal conductivity was not possible to investigate. The thermal properties from KD2 Pro measurements were compared to average diagrams based on Kersten's (1949) empirical equations for thermal conductivity. The comparison showed that thermal conductivity from Kersten's (1949) diagrams are lower than the values obtained by the modern KD2 Pro Thermal Properties Analyzer. Uncertainties are associated with procedure and results for both methods. Hence, more data is needed to argue that the KD2 Pro is a more efficient, accurate and straightforward method to obtain values for thermal conductivity than the average diagrams. Surface and ground temperature data from a thermistor string installed in lower Adventdalen have been collected and plotted with time and depth. In agreement with literature, the attenuation of temperature with depth was shown by a phase lag, or illustrated by a whiplash curve. The seasonal thawed layer in lower Adventdalen is estimated to be approximately one meter in thickness. The mean annual ground temperature is about -5.2\degree C. Back-calculations of thermal diffusivity from temperature data were performed using the thermistor string plots. The approach met some challenges, but the result was in agreement with the direct measurements with the KD2 Pro Thermal Properties Analyzer. Finally, a one-dimensional geothermal model was used to investigate the ground thermal regime response to climate changes. The model was created with a finite element program called PLAXIS and used known thermal and physical properties as input parameters for the soil. The thermal boundary condition at the surface was adapted from present and future air temperature data. The model was run for the period 2016-2017, and the results were validated with sub-surface temperature data from the thermistor sting. The PLAXIS model showed unexpected results regarding the depth of thaw of the upper soil layer. The PLAXIS model overestimated the active layer with about one meter. However, the model showed good agreement with the thermistor string data in deeper layers due to the bottom boundary condition. The same model set-up was used to simulate two future scenarios in 2090; RCP2.6 and RCP8.5. Both climate scenarios indicate a warming of the Earth's surface. Increased thickness of the active layer was seen in both scenarios, with the RCP8.5 initiation the deepest thawing depth. However, conclusions must be considered with great caution due to the simple procedure and many assumptions