Ensemble Simulations of Glacier Retreat over Randomly Generated Self-Similar Subglacial Topography

Simulations of marine-terminating glacier retreat show that the bed topography controls the timing and speed of glacier retreat. Though observations of subglacial topography are difficult to observe at a wide range of scales, radar sounding indicates that subglacial bed topography generally exhibits self-similarity over scales relevant to ice sheet dynamics. To investigate how self-similar topography controls glacier retreat under climate change, we run large ensembles of retreat simulations using a one-dimensional numerical ice-flow model and randomly generated bed topographies. We systematically explore a range of amplitudes and self-similarity characteristics of the subglacial topography across multiple ensembles. We find that topography can have two opposing effects on glacier retreat: large topographic peaks are barriers for retreat (and reduce mass loss in average simulations), while large topographic troughs also result in more mass loss in the most extreme simulations. As the amplitude of topography increases, both effects will occur simultaneously since larger peaks are accompanied by larger troughs. Overall, including larger-amplitude topography leads to greater uncertainty in simulated glacier mass loss (a wider range of outcomes). While our ensemble simulations quantified the effects of topographic variations, there is uncertainty when comparing statistical distributions of uncertain bed topography that must be further explored at various time scales, in addition to increasing the number of ensemble simulations to better capture ensemble statistics. These preliminary experiments will help guide interpretation of simulations, particularly as large ensemble methods and stochastic generation of bed topography become more feasible routes to quantifying uncertainty in future sea level projections.