Abstract by Katie Larson
Shane Reese, William Christensen
Bayesian Computation for High Dimensional Himalayan Glacier Tracking
Glaciers in the Himalayan mountain range provide water for over 1.4 billion people. For years these glaciers have been retreating in response to climate changes in High Mountain Asia. Statistical models can be used to understand this phenomenon and quantify the effects on water resources. Part of the modeling process requires generating samples from complex and high dimensional probability distributions using the Metropolis-Hastings algorithm. The efficiency of this algorithm is sensitive to user-defined choices for values of “tuning" parameters. We review theoretical results proving optimization of the Metropolis-Hastings algorithm in constrained scenarios, and test these results in more realistic conditions. The results from this purely computational research project are being applied in a highly collaborative research setting related to glacier tracking.