Reconciling inferences of mantle viscosity and late Quaternary ice history - the importance of an asthenosphere and 3-D viscosity

Deglaciation history models from the last glacial maximum (LGM) and accompanying mantle viscosity structures have been jointly constrained by fitting paleo-sea level data through glacial isostatic adjustment (GIA) modeling. However, some GIA-inferred 1-dimensional (1-D) viscosity models, including the widely used VM5a, contradict the mantle viscosity structure inferred using other approaches/constraints, such as seismic imaging, mineral physics and mantle convection modeling. Here we show that GIA models with realistic 3-D mantle viscosity derived from seismic structures with an effective activation energy of 300-400 kJ/mol and a weak asthenosphere fit paleo-sea level data in North America better than several widely used 1-D GIA models. Our results also demonstrate that the inclusion of the asthenosphere and lateral viscosity variations permits a larger range of North American ice sheet reconstructions, compared to 1-D models assuming VM5a, that produce quality fits to the sea-level data, including models with substantially greater volume at the LGM, thus identifying a potential solution to balancing the sea level budget during this key climate event. To fit the currently available far-field and near-field data, the North American component of the ICE-6G model must be increased in volume by 10-20%, assuming no changes beyond this region.