Simulations of the Last Interglacial with ICON-XPP indicate the relevance of correct sea-ice and vegetation feedback for Northern Hemisphere warming

The Last Interglacial (LIG; 129–116 thousand years ago) experienced global mean temperatures approximately 1–2 °C above pre-industrial levels, comparable to present-day conditions and those projected for the near future. During the LIG, high and mid-latitudes were substantially warmer, Arctic sea ice was reduced, both the Greenland and Antarctic ice sheets were smaller than today, and global mean sea level was at least 5 m higher than present. Unlike modern warming, which is primarily driven by increased greenhouse gas concentrations, LIG climate anomalies were mainly caused by higher eccentricity and a precession putting NH summer closer to perihelion, with Northern Hemisphere summer insolation exceeding pre-industrial values by more than 70 W m⁻².

 

Many climate models struggle to reproduce the magnitude of LIG warming and the seasonally ice-free Arctic suggested by proxy evidence. Here, we present results from an abrupt-127 ka experiment following the CMIP7 Fast Track protocol, performed with ICON-XPP v1.0 (07/2024) [1], extended to allow for orbital parameter variations based on Kepler’s approximation. In this simulation, orbital parameters and greenhouse gas concentrations are set to LIG values, while all other boundary conditions (solar constant, prescribed ice sheets, prescribed vegetation, and aerosols) are kept at pre-industrial levels.

 

Compared to the pre-industrial control simulation, the abrupt-127 ka experiment shows top-of-atmosphere (TOA) radiation anomalies consistent with previously published LIG simulations [2] including an Arctic summer TOA increase of 50–75 W m⁻². However, for ICON-XPP v1.0 the simulated annual global mean temperature decreases by 0.3 K when only orbital parameters are changed, and by 0.47 K when LIG greenhouse gas concentrations are applied in addition. This contradicts proxy reconstructions indicating a global mean temperature increase of approximately 0.5–1.5 K during the LIG.

 

Exploring Arctic seasonality, we find a summer warming of 4 K in July and a winter cooling of 3 K in January, resulting in an overall Arctic cooling in the annual mean relative to pre-industrial conditions. Arctic sea ice shows little reduction in summer but increases more substantially in winter, leading to an overall annual expansion of sea ice compared to pre-industrial levels. We attribute the simulated cooling and disagreement with proxy evidence to insufficient Arctic amplification in the ICON-XPP version used, likely caused by a weak sea-ice feedback and the lack of interactive vegetation changes. We compare these results to first results obtained with the CMIP7 release of ICON-XPP (2025.10-1) and sensitivity experiments exploring the impact of prescribed vegetation changes and the inclusion of dynamic vegetation. Our findings have major implications for future simulations with ICON-XPP, as the LIG represents a climate state comparable to present-day and future warmth.

 

[1] Müller et al.: The ICON-based Earth System Model for Climate Predictions and Projections (ICON XPP v1.0), EGUsphere, https://doi.org/10.5194/egusphere-2025-2473, 2025.

[2] Otto-Bliesner et al.: Large-scale features of Last Interglacial climate: results from evaluating the lig127k simulations for the Coupled Model Intercomparison Project (CMIP6)–Paleoclimate Modeling Intercomparison Project (PMIP4), Clim. Past, https://doi.org/10.5194/p-17-63-2021, 2021.