07-12, 15:10–15:20 (Europe/Amsterdam), While Loop (4.2)
Modeling subglacial water flow and its link to ice dynamics is necessary for accurate predictions of ice sheet response to a warming climate. Here we present a re-implementation of the widely used GlaDS model to run on GPUs. We show-case the matrix free implementation which leverages the full capabilities of the GPU, present model runs of test cases, show the model's scalability and provide an outlook towards inversion schemes and high-resolution continental-scale applications.
Subglacial water flow plays an important role in the dynamic of a glacier or ice sheet by impacting the conditions at their base and thus the sliding. Understanding subglacial drainage and its links to ice flow is necessary to predict sea level rise due to the ice sheets' evolution in a warming climate.
The Glacier Drainage System Model (GlaDS), one of the most widely used such models, simulates both channelised and distributed drainage at the ice-bed interface. Here we present a re-implementation of GlaDS running on Graphical Processing Units (GPUs). The aim is for the model to run on meshes larger than 10,000² grid points, which would allow, for instance, to simulate Antarctica at 500m resolution. Unlike the original GlaDS implementation, this is based on a finite difference scheme on a structured grid. Together with a matrix-free solver, this allows us to leverage the full performance capabilities of GPUs. We present model runs of test cases, show the model's scalability and provide an outlook towards inversion schemes and higher-resolution continental-scale applications.
Geo-HPC, Julia GPU & Supercomputing.
I am a glaciologist and Julia programmer. I work at ETH-Zurich and the Swiss Federal Institute for Forest, Snow and Landscape Research (WSL).
