Fundamentals, accuracy and input parameters of frost heave prediction models

In this thesis, the frost heave knowledge of physical geographers and soil physicists, a detailed description of the frost heave process, methods to determine soil parameters, and analysis of the spatial variability of these soil parameters are connected to the expertise of civil engineers and mathematicians in the (computer) modelling of the process. A description is given of observations of frost heave in laboratory experiments and in the field. Frost heave modelling is made accessible by a detailed description of the main principles of frost heave modelling in a language which can be understood by persons who do not have a thorough mathematical background. Two examples of practical one-dimensional frost heave prediction models are described: a model developed by Wang (1994) and a model developed by Nixon (1991). Advantages, limitations and some improvements of these models are described.; It is suggested that conventional frost heave prediction using estimated extreme input parameters may be improved by using locally measured input parameters. The importance of accurate input parameters in frost heave prediction models is demonstrated in a case study using the frost heave models developed by Wang and Nixon. Methods to determine the input parameters are discussed, concluding with a suite of methods, some of which are new, to determine the input parameters of frost heave prediction models from very basic grain size parameters. The spatial variability of the required input parameters is analysed using data obtained along the Norman Wells-Zama oil pipeline at Norman Wells, NWT, located in the transition between discontinuous and continuous permafrost regions at the northern end of Canada's northernmost oil pipeline. A method based on spatial variability analysis of the input parameters in frost heave models is suggested to optimize the improvement that arises from adequate sampling, while minimizing the costs of obtaining field data.; A series of frost heave predictions is made using a modified version of the model of Wang and the determined series of input data along the Norman Wells pipeline. The spatial variation in computed frost heave, an indicator of differential frost heave resulting from a spatial variation of input parameters, is discussed. The thesis concludes with an analysis of the sources of potential errors in, and an evaluation of the merits of frost heave prediction.