Science [&] Technology Norway (NO)
The instability of Antarctic ice shelves is one of the most critical open questions in polar science, due to its capacity to drive rapid sea level change at – and beyond – current high-end climate projections. Yet forecasting future instability is notoriously difficult because of the complex, non-linear forcing mechanisms controlling an ice shelf’s response, including calving at the ice margin. As a result, the timescales for ice shelf collapse forms one of the largest uncertainties in modelling future sea level scenarios.
In AI4IS we aim to develop the first AI-based forecasting system for iceberg calving of Antarctic ice shelves. Our AI model, which will fundamentally be built to include Explainable AI (XAI) techniques, will consume a bespoke 4-D multivariate data cube of EO products, complemented with process-model simulations of key climate parameters. Our 4-D data cube will be assembled using a novel Gaussian random field representation approach that our team have recently developed, which is computationally efficient and preserves sub-grid scale information.
