Coupling coincident satellite observations and machine learning to improve ice-sheet models and sea-level projections

Approximately 3% of the global population (230 M) live within 1 m of current sea level (Kulp & Strauss 2019). These low-lying coastal areas are also home to many forms of vital infrastructure such as transport networks, power stations and factories. However, these areas are increasingly at risk of catastrophic flooding as projections of global sea-level rise suggest that oceans may rise by up to 1 m by 2100 (IPCC AR6 2021).

Furthermore, these predictions have large uncertainties, in part due to the uncertain contributions to future sea level from the Antarctic ice sheet. Accurate and reliable ice-sheet models are vital for reducing uncertainties in projections of future global sea-level.

Uncertainties can be reduced by improving our understanding of complex ice-sheet processes, such as ice damage and ice-ocean interactions, which are crucial for assessing ice-sheet stability. The ever increasing volume of satellite observations from Antarctica signals a need to combine and distill information from multiple sensors so that they can be fully utilised to investigate ice-sheet dynamics.

We propose to harness multiple sets of remote-sensing satellite observations and develop a digital infrastructure to process and visualise coincident observations of ice-sheet surface imagery and elevation in Antarctica. This new dataset will be the first of its kind and a valuable resource for improving our understanding of ice-sheet change. It will be made freely available as a resource for the scientific community and citizen scientists.

Furthermore, as part of the project we will use this new dataset to assess the evolution of ice-sheet damage using machine learning thereby demonstrating the advantages of assessing multiple forms of satellite data together.

This project will embed a glaciologist from the University of Edinburgh with expertise in geophysical data analysis and ice-sheet modelling within the operations of earth-observation specialists, EarthWave. The aims of the project are:

- To increase the digital capabilities of EarthWave to handle satellite imagery as part of their existing multi-satellite data service.

- To make the Antarctic multi-satellite data product freely available, suitable for, and of interest to, the wider scientific community. - To generate a dataset of ice-surface imagery and elevation to investigate ice-sheet damage. - Train a neural network to quantify ice damage from satellite imagery.

Direct Stakeholders: Host Organisation: EarthWave, Embedded Research: M. Wearing,

Wider Stakeholders: NERC, NERC-NSF ITGC, ITGC PROPHET, National and Local government, European Space Agency, NASA, General Public/Citizen Scientists, Coastal Communities

Keywords: Satellite remote-sensing, earth observation, sea-level rise, Antarctica, big-data, artificial intelligence, machine learning, glaciology, ice sheets, ice shelves