When the research for this study started (2015), I had been using Julia as my main programming language for two years. I knew Julia would be suitable for this project because of its high performance and ease of use.
Mauro Werder
ETH
Melting glaciers and melting ice caps are responsible for a large share of projected sea level rise over the next several decades. In fact, these fresh water sources are expected to contribute more to rising sea levels during the first half of this century than the melting ice sheets of Greenland and Antarctica.
In order to accurately project sea level rise, glaciologists and other climate scientists require accurate maps of ice thickness.
Glaciologist Mauro Werder and his co-authors Matthias Huss, Frank Paul and Amaury Dehecq use Julia as part of a new method to infer ice thickness maps of glaciers. They estimated maps of ice thickness for more than 30,000 glaciers around the globe.
Mauro explains:
“When the research for this study started (2015), I had been using Julia as my main programming language for two years. I knew Julia would be suitable for this project because of its high performance and ease of use.
For this study, performance was critical. We used a Bayesian approach implemented with a Markov chain Monte Carlo (MCMC) method which necessitated millions of model evaluations. We calculated more than 100 million ice thickness maps. Julia seamlessly provides this performance without having to jump through hoops. For example, if we had used Python, we would have had to use the restrictive Numba JIT, or use another language.
Equally important, Julia is a joy to program in and very productive.
It’s simply much easier to write and combine performance critical and uncritical code in Julia compared with other languages such as Matlab, Python or Fortran. Julia excels at both and doesn't require any code acrobatics as other languages often require.
As a result, we created a new method to infer ice thickness maps of glaciers. We used and enhanced an existing forward model and estimated its parameters and outputs using a Bayesian approach. We estimated maps of ice thickness for more than 30,000 glaciers from around the globe. This type of exercise has never before been conducted with such a rigorous analysis of the uncertainties of the estimates."
A Bayesian Ice Thickness Estimation Model for Large-Scale Applications is available here and the code-repo is available on GitHub.
