Antarctic ice sheet mass loss, glacio-isostatic adjustment and surface processes from a Bayesian combination of gravimetry, altimetry and GPS data
Abstract:Constraining past ice mass changes, identifying their cause(s) and determining rigorous error estimates, is important for closing the sea level budget and as an input for and test of numerical models. Despite the progress that has been made over the last decade, significant differences remain for estimates of the mass evolution of the Antarctic ice sheet. These estimates often yield conflicting results with non-overlapping error bars, while the commonly adopted use of different forward models to isolate and remove the effects of glacio-isostatic adjustment (GIA) and surface mass balance (SMB) processes introduces another source of uncertainty which is hard to quantify. To address both these issues, we present a statistical modeling approach that utilises a spatio-temporal Bayesian hierarchical model, alongside novel dimensional reduction methods to allow the solution to remain tractable in the presence of the large number (> 10^7) of observations. We solve simultaneously for GIA, surface processes, elastic rebound, firn compaction and ice dynamics.
Over 2003-2013, Antarctica has been losing mass at a rate of -82+-23 Gt/yr. West Antarctica is the largest contributor with -114+-10 Gt/yr, mainly triggered by high thinning rates of glaciers draining into the Amundsen Sea Embayment. The Antarctic Peninsula has experienced a dramatic increase in mass loss in the last decade, with a mean rate of -25+-6 Gt/yr, and significantly higher values for the most recent years following the destabilization of the Southern Antarctic Peninsula around 2010. The total mass loss is partly compensated by a significant mass gain of 57+-20 Gt/yr in East Antarctica due to positive SMB anomalies and an interesting small dynamic component. We compare our time series of SMB anomalies with those from RACMO-2.3, obtaining good agreement for the large-scale patterns, although differences arise at a basin scale. Also, a data-driven GIA solution is obtained which could be used to constrain and validate existing and future forward models.