Item

Abstract

Accurately projecting mass loss from ice sheets is critical to help societies best prepare for the change in sea level. Despite tremendous improvements, several recent studies show that the agreement between models and the observational record remains poor. The inability of numerical models to reproduce observations raises concerns about their ability to provide accurate projections. Data assimilation approaches are great tools to infer unknown parameters by minimizing the misfit between model and observations. Inversions have been used in glaciology since the 1990s, but only for a given point in time. These “snapshot inversions” are routinely used to infer unknown parameters, such as basal friction, but they do not take advantage of time series of observations to which we have access today. The advent of Automatic Differentiation and its recent integration in the Ice-sheet and Sea-level System Model and STREAMICE makes it possible to assimilate almost any type of data using time dependent models. Here we apply transient calibration to the Amundsen Sea Embayment between 2004 and 2022 using surface velocities from MEaSUREs and ITS_LIVE, and surface altimetry data from Cryosat and ICESat-2. We assess the performance of transient compared to snapshot calibrations in terms of capturing past and current trends in speed change, thinning, grounding line retreat and mass change. We then compare future projections over the next 100 years. This exercise paves the way to future modeling work that makes use of more dense time series to constrain critical model parameters and reduce uncertainty in future sea level rise.