The replication of the behaviour of frozen soils has been studied for decades. Many attempts have been undertaken to either develop new constitutive models or to improve already existing models to simulate the behaviour of frozen materials. To handle the challenges of ground freezing, cold regions engineering and periglacial processes, it is vital to understand the mechanical behaviour of frozen soil. Knowing that field studies, large scale laboratory tests and centrifuge modelling offer good insights, they are however expensive and time consuming activities to undertake. A numerical modelling approach is therefore necessary. The Norwegian University of Science and Technology (NTNU), in collaboration with Plaxis bv, developed a new numerical model to tackle the afore-mentioned problems. The aim of this new approach is to provide a reliable design tool to assess the impact of climate change and changes in temperature in general on a variety of engineering problems. The constitutive model requires several parameters of which quite a few are not very common to geotechnical engineers. Furthermore, to analyse frozen soil and the behaviour when phase transition occurs, specific properties have to be taken into account which are not determined in standard site investigation and soil lab testing campaigns. This brings the need for a simplified method to determine such properties based on data that is commonly available, such as the particle size distribution. The idea is, as an initial estimate, to correlate this data to the soil freezing characteristic curve and the hydraulic properties of partially frozen soils. Therefore, a practical approach to obtain crucial properties of frozen soil such as the soil freezing characteristic curve (SFCC), the freezing/melting point of a soil-water system and its hydraulic conductivity by means of limited input data is developed. Different models and empirical equations are combined to provide a closed formulation which can be used in computer simulations to account for moisture migration in partially frozen soils. Input data such as grain size distribution and dry bulk density suffices to obtain the aforementioned properties. Further consideration of the pressure dependence of the freezing/melting temperature of water/ice even allows accounting for the phase change point depression and thus the phenomenon of pressure melting. The model is appropriate not just to represent qualitatively the SFCC of different soil types, but also to provide conformity between the model prediction and measured data of many soil types having a log-normal grain size distribution. This user-friendly approach is tested in the geotechnical finite element code PLAXIS 2D. Although some correlations and default values can be provided, laboratory testing is inevitable in order to provide the complete set of necessary soil parameters. To calibrate some of the most difficult ones, namely the Barcelona Basic Model (BBM) parameters, the idea is to use oedometer test results. The temperature-controlled oedometer test requires equipment which is less sophisticated than a temperature-controlled isotropic compression test. A shorter testing period makes it possible to save time and money. An optimisation approach for identification of material parameters in elasto-plastic models for unsaturated soils, like the BBM, using the results from suction-controlled oedometer tests is developed by Zhang et al. (2016). The same approach, reformulated for the constitutive model for frozen and unfrozen soil, is explained in this thesis. This optimisation approach allows to simultaneously determine parameters governing isotropic virgin behaviour as well as unloading and reloading behaviour. A number of simplified real life applications are presented in this thesis. They demonstrate the correctness of the model's theoretical implementation and its correct representation of the behaviour of frozen soil found in nature. Temperature gradients under uniaxial compression are applied on a frozen sample, and show that the hardening modulus and the uniaxial compression strength increase as the average temperature and the thermal gradient decrease. Another boundary value problem investigates the pressure melting phenomenon and shows that high confining pressures cause the melting of ice crystals due to the reduction of the phase transition temperature. Two freeze-thaw cycles of a clay sample are simulated and provide information on the ability of the model to simulate the ice segregation phenomenon (frost heave) as well as the thaw settlement behaviour. Furthermore, three more practical applications are presented. First, a pipeline is installed in an unfrozen trench and subsequently cooled down by the streaming of a chilled fluid. The onset of frost heave may potentially cause many engineering problems, like cracking of pavements and the fracturing of the pipeline. It is therefore of particular concern in highway and pipeline engineering. The second application comprises a foundation on frozen soil subjected to a warming period. Settlements due to the thawing of ice enclosed in the frozen soil layers accumulate and should be considered when designing embankments and building foundations on permafrost. Finally, an application of artificial ground freezing is considered. A tunnel is constructed with the use of freeze pipes in order to stabilise the soil during excavation. The soil is frozen by means of the installed freeze pipes. Watertightness and an increase in strength of the soil are achieved. Once the soil has frozen sufficiently, tunnel construction can take place. To sum up, many essential features of the mechanical behaviour of frozen and unfrozen soil can be captured with this new constitutive model. The dependence of stiffness and shear strength on temperature are among of these. Furthermore, two main features, namely frost heave and thaw settlements, can be simulated. These phenomena play a key role in designing in, on and with frozen / unfrozen soil and may cause significant engineering problems. Three conference papers for two different conferences have been submitted and reflect my research results. However, all that glitters is not gold. The investigated constitutive model cannot fully capture all the effects influencing the behaviour of frozen soil. The non-incorporation of cyclic and time-dependent behaviour represents one of a few shortcomings.