Probabilistic analysis of thaw penetration using global climate models in Fairbanks, Alaska

Abstract

The depth to which freezing and thawing may occur in soils is important to the design of pavements, embankments, structures and utilities in areas of seasonal frost and permafrost. In the prediction equations the material property and surface energy input parameters (e.g. air temperature, cloud cover, snow cover) are often assumed to be deterministic. In reality the input parameters are not fixed values but random variables. A probabilistic analysis may be performed using a deterministic model and the probability density function for the input parameters. To illustrate the probabilistic approach, an example of the prediction of thaw depths in Fairbanks, Alaska, is considered. More specifically, the Stefan equation and modified Berggren equation are used together with the Monte Carlo simulation technique to make a probabilistic prediction of thaw penetration. The simulation results indicate that the variability in material properties, surface energy input parameters and temperature data can lead to significant uncertainty in predicting thaw penetration. With respect to determining the mean and standard deviation of thaw penetration, Monte Carlo simulation and Taylor's series method were performed and compared. The Taylor's series method also offered results close to that calculated by Monte Carlo simulation. The close agreement of the results suggest the Taylor's series method may be applied to solve many cold region's engineering problems to account for the variability of input parameters.