Dynamics of Ice Streams: A Physical Statistical Approach

Ice streams are believed to play a major role in determining the response of
their parent ice sheet to climate change, and in determining global sea level
by serving as regulators on the fresh water stored in the ice sheets. Ice
streams are characterized by rapid, laterally confined flow which makes them
uniquely identifiable within the body of the more slowly and more homogeneously
flowing ice sheet. But while these characteristics enable the identification of
ice streams, the processes which control ice-stream motion and evolution, and
differences among ice streams in the polar regions, are only partially
understood. Understanding the relative importance of lateral and basal drags,
as well as the role of gradients in longitudinal stress, is essential for
developing models for future evolution of the polar ice sheets. In this
project, physical statistical models are used to explore the processes that
control ice-stream flow, and to compare these processes between seemingly
different ice-stream systems. In particular, the Northeast Ice Stream in
Greenland will be investigated. Geophysical models lie at the core of the
approach, but are embellished by statistical modeling of various components of
variability. One important component comes from the uncertainty in
observations on basal elevation, surface elevation, and surface velocity. In
this project, new observational data collected using remote-sensing techniques
are used. The various components, some of which are spatial, are combined
hierarchically using Bayesian statistical methodology. All these are combined
mathematically into a physical statistical model that yields the posterior
distributions for basal and surface elevations, surface velocity fields, and
stress fields, conditional on the data. Inference based on these distributions
is carried out via Markov chain Monte Carlo techniques, to obtain estimates of
these unknown fields along with uncertainty measures associated with them.