Greenland warming during the last interglacial : the relative importance of insolation and oceanic changes

Insolation changes during the Eemian (the last interglacial period, 129–116,000 years before present) resulted in warmer than present conditions in the Arctic region. The NEEM ice core record suggests warming of 8±4 K in northwestern Greenland based on water stable isotopes. Here we use general circulation model experiments to investigate the causes of the Eemian warming in Greenland. Simulations of the atmospheric response to combinations of Eemian insolation and preindustrial oceanic conditions and vice versa, are used to disentangle the impacts of the insolation change and the related 5 changes in sea surface temperatures and sea ice conditions. The changed oceanic conditions cause warming throughout the year, prolonging the impact of the summertime insolation increase. Consequently, the oceanic conditions cause annual mean warming of 2 K at the NEEM site, whereas the insolation alone causes an insignificant change. Taking the precipitation changes into account, however, the insolation and oceanic changes cause more comparable increases in the precipitationweighted temperature, implying that both contributions are important for the ice core record at the NEEM site. The simulated 10 Eemian precipitation-weighted warming of 2.4 K at the NEEM site is low compared to the ice core reconstruction, partially due to missing feedbacks related to ice sheet changes. Surface mass balance calculations with an energy balance model indicate potential mass loss in the north and southwestern parts of the ice sheet. The oceanic conditions favor increased accumulation in the southeast, while the insolation appears to be the dominant cause of the expected ice sheet reduction.

The conversion from δ 18 O to temperature may be a contributing factor to mismatches between model simulations and δ 18 O temperature reconstructions: The δ 18 O-temperature relationship is sensitive to precipitation intermittency, evaporation conditions, and atmospheric transport, and thus varies spatially and historically (Jouzel et al., 1997;Masson-Delmotte et al., 2011).Hence, sea surface warming and reduced sea ice extent might thus affect the δ 18 O record, as illustrated by isotopeenabled climate model simulations (Sime et al., 2013).The NEEM δ 18 O temperature estimate is based on the average Holocene δ 18 O-temperature relationship from other central Greenland ice cores (Vinther et al., 2009), but the actual relationship might be different due to the shifted location or the climatic changes during the Eemian.
The experiments and the employed models are described in Sect. 2. Results are presented and discussed in Sect.3, followed by conclusions in Sect. 4.
The insolation is internally calculated following Berger (1978) using the same code modification as Muschitiello et al. (2015).
Due to the diverse ice sheet reconstructions, we have kept the ice sheets fixed at present day extents in all of our simulations.

Surface mass balance calculations
To investigate the impacts of the simulated climate changes on the GrIS surface mass balance, we performed off-line calculations with the subsurface scheme of the HIRHAM5 regional climate model (updated from Langen et al. (2015)) :::::::::::::::::::::::::::: (updated from Langen et al., 2015).The subsurface model was here run on the Gaussian grid associated with the EC-Earth experiments and forced at 6 hour intervals with incoming shortwave and downward longwave radiation, latent and sensible heat fluxes, along with rain, snow and evaporation/sublimation taken directly from the EC-Earth output.The subsurface model was updated slightly compared to that described by Langen et al. (2015); most notably it employs 25 layers with a total depth of 70 m water equivalent and includes temperature-and pressure-dependent densification of snow and firn (following Vionnet et al. (2012)) ::::::::::::::::::::::::: (following Vionnet et al., 2012).It accounts for heat diffusion, vertical water transport and refreezing.
Each layer can hold liquid water corresponding to 2 % of the snow pore space volume and excess water percolates downward to the next layer.When a layer density exceeds the pore close off density (830 kg m −3 ), water percolating down from above is added to a slush layer and runs off exponentially with an exponential time scale depending on surface slope (Lefebre et al., 2003;Zuo and Oerlemans, 1996).Until it runs off, the slush layer water is available for superimposed ice formation onto the underlying ice layer at a rate that assumes a linear temperature profile in that layer.
A small area in northwestern Greenland is warming through winter and spring (March-April-May; MAM).Again, this warming coincides with loss of snow cover, and increased sensible heat release from the surface.Continental snow cover changes can thus extend the summertime warming to the colder seasons with reduced insolation, but the iL+oP experiment indicates that this memory effect only plays a minor role for GrIS as a whole.

Precipitation-weighted temperature
The precipitation changes in Fig. 4 suggest that the northwestern GrIS near the NEEM ice core location is affected by changed precipitation seasonality in all three simulations: the insolation in iL+oP causes increased summer snowfall and drier conditions in fall, the oceanic changes in iP+oL cause increased snowfall throughout the year, and the combination in iL+oL leads to increased snowfall during spring and summer.To assess how these changes might affect the ice core record, the precipitationweighted annual mean temperature has been calculated following Eq.(1); Figure 5 compares the annual mean temperature change and the precipitation-weighted mean.

Table 2 .
Insolation and GHGs SSTs and sea ice Annual mean (Tann) and precipitation-weighted (Tpw) temperature change relative to iP+oL and associated standard deviations (σann, σpw) for dNEEM.* Not statistically significant at the 95 % confidence level.