Calibrated Mass Loss Predictions for the Greenland Ice Sheet.

The potential contribution of ice sheets remains the largest source of uncertainty in predicting sea‐level due to the limited predictive skill of numerical ice sheet models, yet effective planning necessitates that these predictions are credible and accompanied by a defensible assessment of uncertainty. While the use of large ensembles of simulations allows probabilistic assessments, there is no guarantee that these simulations are aligned with observations. Here, we present a probabilistic prediction of 21st century mass loss from the Greenland Ice Sheet calibrated with observations of surface speeds and mass change using a novel two‐stage surrogate‐based approach. Our results suggest a sea‐level contribution ranging from 4 to 30 cm at the year 2100, proviso the assumption that our chosen ice sheet model's physics represent reality. Plain Language Summary: Predicting sea level rise is important for social planning, but the contributions to sea level rise from ice sheets are still very uncertain. Contributions from ice sheets are uncertain because they behave according to difficult to observe physics that we're still a little bit fuzzy on, and this incomplete knowledge makes its way into the models that we use to predict the future of the ice sheets. To account for this and to simulate the full range of possible futures for the Greenland Ice Sheet, we run a large number of model simulations where we vary parameters that are not well known. We then pare these down based on which ones agree with some data collected by satellites. Based on these results, we think that the Greenland Ice Sheet could contribute between 2 inches and a foot to sea level rise by the year 2100, depending on what people do about greenhouse gas emissions. Key Points: Credible predictions of glacier mass loss need to be conditioned on observationsWe present a two‐stage Bayesian calibration to condition ensemble predictions on observationsConditioning on observations of surface speeds and cumulative mass loss corrects bias and reduces the credibility interval [ABSTRACT FROM AUTHOR]