Disentangling Natural Versus Anthropogenically-Forced Retreat of Marine-Terminating Glaciers

Many marine-terminating outlet glaciers in Greenland and Antarctica have retreated in recent decades, but the role of anthropogenic forcing is uncertain. One challenge is that internal climate variability obscures anthropogenic trends in the regional atmospheric and ocean conditions that affect glaciers. Recent studies have made progress in partitioning natural and anthropogenic components of these forcing anomalies (e.g., Holland et al., 2019), but formal attribution statements have not yet been made specifically for marine-terminating glacier retreat. A compounding challenge for attribution stems from a glaciers dynamical response to forcing. Marine-terminating glaciers on reverse-sloping beds may retreat irreversibly if sufficiently perturbed -- a process known as the marine-ice-sheet instability. Indeed, many termini or grounding lines have been observed to retreat rapidly from local bed peaks. The combination of internal climate variability and glacier instabilities thus raises the question as to whether natural variability alone might exceed the threshold for triggering rapid retreat. Here, we propose a probabilistic attribution framework to address these challenges using large ensembles of glacier models forced with independent realizations of climate variability. The approach is to estimate the likelihood that climate variability alone could trigger retreat, and the extent to which anthropogenic forcing changes that likelihood. Synthetic experiments over a range of parameters clarify the leading controls on the probability of retreat. We show that century scale trends in external forcing robustly increase the probability of retreat, even if the trend is small compared to the natural variability. This is a consequence of the timescales over which marine-terminating glaciers integrate forcing, and suggests that the full history of anthropogenic forcing is important for attribution, even for recently initiated retreats.