Detecting Sea Level Fingerprints from Synthetic Satellite Altimetry Data Using Deep Learning

Continental freshwater input from glaciers and ice sheets is responsible for more than half of the ongoing global sea level rise. This freshwater redistributes across the oceans following specific patterns, determined by gravitational, rotational and deformation effects, known as sea level fingerprints. These fingerprints can be uniquely associated with their continental mass sources and could in theory enable the reconstruction of continental water and ice mass changes, helping to better attribute the causes of ongoing sea level change. However, they are very difficult to detect because their magnitude is much smaller than the signals related to ocean sterodynamic changes and atmospheric effects. To address this challenge, our research has employed deep learning techniques to separate sea level fingerprints from synthetic satellite altimetry data. Our findings reveal that deep learning is highly effective at this task, highlighting significant potential of deep learning in detecting large-scale geospatial signals. This deep learning approach could serve as a basis for accurately quantifying mass changes in the cryosphere and land hydrology from satellite altimetry observations over the last three decades, ultimately providing valuable insights into the impacts of climate change on sea level and the global water cycle.