Videos from JuliaCon are now available online
2018
Previous editions: 2017 | 2016 | 2015 | 2014
Mauro Werder

ETH Zurich, Switzerland



BITE, a Baysian Ice Thickness Estimation model

The around 200’000 glaciers and ice caps around the world are projected to dominate global sea level rise up to 2100. Whilst air-borne and ground-based ice penetrating radar works very well, it cannot be applied to all glaciers. Thus to estimate their thickness methods based on remotely sensed surface data are needed. We present such a method based on a simple, semi-physical forward model based on mass conservation and basic ice physics and combine it with a Bayesian inversion approach. Using MCMC techniques we can estimate maps of ice thickness and their errors. The forward model is reasonably well tuned and can calculate a thickness map of a big glacier in less than 0.1s. The inversion is done using KissMCMC.jl which implements a Affine Invariant Markov chain Monte Carlo (MCMC) Ensemble sampler modeled after the Python emcee implementation.

Speaker's bio