A note on the influence of atmospheric model resolution in coupled
climate–ice-sheet simulations
Abstract. Coupled climate–ice-sheet simulations have been growing in popularity in recent years. Experiments of this type are however challenging as ice sheets evolve over multi-millennial time scales, which is beyond the practical integration limit for most Earth-system models. A common method to increase model throughput is to trade resolution for computational efficiency (compromises accuracy for speed). Here, we analyze how the resolution of an atmospheric general circulation model (AGCM) influences the simulation quality of a standalone ice-sheet model. Four identical AGCM simulations of the Last Glacial Maximum (LGM) were run at different horizontal resolutions: T85 (1.4°), T42 (2.8°), T31 (3.8°), and T21 (5.6°). These simulations were subsequently used as forcing of an ice-sheet model. While the T85 climate forcing reproduces the LGM ice sheets to a high accuracy, the intermediate resolution cases (T42 and T31) fail to build the Eurasian Ice Sheet. The T21 case fails in both Eurasia and North America. Sensitivity experiments using different surface mass balance parameterizations improve the simulations of the Eurasian ice-sheet in the T42 case, but the compromise is a substantial ice buildup in Siberia. The T31 and T21 cases are not improving in the same way in Eurasia, though the latter simulates the continent-wide Laurentide Ice Sheet in North America. The difficulty to reproduce the LGM ice sheets in the T21 case is in broad agreement with previous studies using low-resolution atmospheric models, and is caused by a deterioration of the atmospheric climate between the T31 and T21 resolutions. It is speculated that this deficiency may demonstrate a fundamental problem using low-resolution atmospheric models in these types of experiments.