Sensitivity of the last glacial inception to initial and boundary conditions: lessons from a coupled climate-ice sheet model

Proxy data indicate that the last glacial inception started at approximately 115ka BP when boreal summer insolation reached its minimum. At that time, large ice sheets started to form in Northern Canada. A number of models of different complexities have been employed to simulate the last glacial inception; however, complex climate models did not incorporate interactive ice sheets. Here, a state-of-art Earth System Model AWI-ESM-2.2 (Gierz et al., 2020, GMDD), composed of AWI-ESM-2.0 (Sidorenko et al., 2019) that now includes the Parallel Ice Sheet Model PISM (The PISM authors, 2016), was utilized to study the potential causes of the inception. By conducting different sensitivity experiments, we investigated the effect of initial conditions, different surface mass balance schemes, greenhouse gas (GHG) concentration, ocean circulation, and model resolution on the last glacial inception.

Two experiments were conducted under 115 ka BP orbital and radiative forcing to examine the effect of initial conditions: one without interactive ice sheets but with fixed preindustrial topography, and the other with interactive ice sheets that include initial snow cover over two small regions in northeastern and northwestern Canada. The first experiment failed to produce a permanent appearance of snow over North America. The second experiment simulated a further growth of ice sheets over northeastern and northwestern Canada. In these experiments, the initial ice sheets provide important feedbacks to cause North America ice sheet growth: the snow-albedo feedback and elevation effect of orography reinforce the cooling in the region initially covered by snow or ice.

We compared an empirically-based positive-degree-day (PDD) scheme, which estimates surface melt as a function of temperature, with a more physically-based diurnal energy balance model (dEBM) (Krebs-Kanzow et al., 2018), which also takes changes in shortwave radiation into account and implicitly resolves a diurnal freeze-melt cycle. Both simulations showed a tendency of ice sheet growth in Northern Canada ice sheet. The experiment employing the dEBM model for surface mass balance had a larger magnitude in SMB, resulting in faster development of the ice sheet.

Another experiment with a lowered GHG concentration was carried out to investigate the role of GHG, suggesting that GHG changes also contribute to a cooling state. Two additional experiments also explored the effect of changed ocean circulation and atmosphere dynamics and their contribution to the inception, as well as the effect of an improved resolution in the atmosphere model.

In summary, our findings imply that initial conditions have a significant impact on simulating the inception. We conclude that the incorporation of an ice sheet model into the Earth system is an important step forward to provide a more realistic simulation of glacial inception and to uncover its potential causes.