Statistical problems with Alpine permafrost
Anthony Davison (EPFL Lausanne)
Monday 30th April, 2012 16:00-17:00 Maths 203
Permafrost consists of soil and rock that remains at sub-zero temperatures for at least 24 successive months. It is an important phenomenon at high latitudes and high altitudes, both because its presence affects infrastructure (oil pipelines, ski-lifts etc.) and because it is sensitive to climate change. Colleagues at the Institute for Snow and Avalanche Research in Davos monitor below-ground temperatures at a number of locations in the Alps, and produce fairly long time series at different depths within boreholes. This talk will describe two statistical problems arising from these data: modelling of the temperature field based on these time series; and the detection of non-conduction at certain depths and times. The first involves fitting a discretized version of the heat equation using MCMC simulation, with the boundary conditions estimated statistically. The second involves order-restricted inference based on parallel periodograms. The work is joint with Juliette Blanchet, Marcia Phillips, and Evelyn Zenklusen Mutter.