Assessment of the Greenland ice sheet-atmosphere feedbacks for the next century with a regional atmospheric model coupled to an ice sheet model

In the context of global warming, growing attention is paid to the evolution of the Greenland ice sheet (GrIS) and its contribution to sea-level rise at the centennial timescale. Atmosphere-GrIS interactions, such as the temperature-elevation and the albedo feedbacks, have the potential to modify the surface energy balance and thus to impact the GrIS surface mass balance (SMB). In turn, changes in the geometrical features of the ice sheet may alter both the climate and the ice dynamics governing the ice sheet evolution. However, changes in ice sheet geometry are generally not explicitly accounted for when simulating atmospheric changes over the Greenland ice sheet in the future. To account for ice sheet-climate interactions, we developed the first two-way synchronously coupled model between a regional atmospheric model (MAR) and a 3-D ice sheet model (GRISLI). Using this novel model, we simulate the ice sheet evolution from 2000 to 2150 under a prolonged representative concentration pathway scenario, RCP8.5. Changes in surface elevation and ice sheet extent simulated by GRISLI have a direct impact on the climate simulated by MAR. They are fed to MAR from 2020 onwards, i.e. when changes in SMB produce significant topography changes in GRISLI. We further assess the importance of the atmosphere-ice sheet feedbacks through the comparison of the two-way coupled experiment with two other simulations based on simpler coupling strategies: (i) a one-way coupling with no consideration of any change in ice sheet geometry; (ii) an alternative oneway coupling in which the elevation change feedbacks are parameterized in the ice sheet model (from 2020 onwards) without taking into account the changes in ice sheet topography in the atmospheric model. The two-way coupled experiment simulates an important increase in surface melt below 2000m of elevation, resulting in an important SMB reduction in 2150 and a shift of the equilibrium line towards elevations as high as 2500 m, despite a slight increase in SMB over the central plateau due to enhanced snowfall. In relation with these SMB changes, modifications of ice sheet geometry favour ice flux convergence towards the margins, with an increase in ice velocities in the GrIS interior due to increased surface slopes and a decrease in ice velocities at the margins due to decreasing ice thickness. This convergence counteracts the SMB signal in these areas. In the two-way coupling, the SMB is also influenced by changes in finescale atmospheric dynamical processes, such as the increase in katabatic winds from central to marginal regions induced by increased surface slopes. Altogether, the GrIS contribution to sea-level rise, inferred from variations in ice volume above floatation, is equal to 20.4 cm in 2150. The comparison between the coupled and the two uncoupled experiments suggests that the effect of the different feedbacks is amplified over time with the most important feedbacks being the SMB-elevation feedbacks. As a result, the experiment with parameterized SMB-elevation feedback provides a sea-level contribution from GrIS in 2150 only 2.5% lower than the two-way coupled experiment, while the experiment with no feedback is 9.3% lower. The change in the ablation area in the two-way coupled experiment is much larger than those provided by the two simplest methods, with an underestimation of 11.7% (14 %) with parameterized feedbacks (no feedback). In addition, we quantify that computing the GrIS contribution to sea-level rise from SMB changes only over a fixed ice sheet mask leads to an overestimation of ice loss of at least 6% compared to the use of a time variable ice sheet mask. Finally, our results suggest that ice-loss estimations diverge when using the different coupling strategies, with differences from the two-way method becoming significant at the end of the 21st century. In particular, even if averaged over the whole GrIS the climatic and ice sheet fields are relatively similar; at the local and regional scale there are important differences, highlighting the importance of correctly representing the interactions when interested in basin scale changes.