The North American Late Wisconsin ice sheet and mantle viscosity from glacial rebound analyses

Observations of sea level and crustal response to glacial loading cycles provide constraints on the mantle rheology function, E, and as well as on the ice load, I, with the latter being largely free from a-priori glaciological or climate assumptions and appropriate, therefore, for testing any such hypotheses. This paper presents new results for both continental-mantle E and I for the Late Wisconsin ice sheet, using geological evidence for relative sea-level change (rsl) and tilting of palaeo-lake shorelines, complemented with loose constraints from observations of present-day radial crustal displacement across North America. The focus is on evidence from near or within the former maximum ice margins and the resulting earth response is representative of sub-continental mantle conditions. The inversion of the sea-level information has limited resolution for earth rheology and simple three-layer models, characterized by depth-averaged effective lithospheric thickness (H) and upper- and lower-mantle viscosities (ηum and ηum respectively) adequately describe the response function, yielding parameters (earth model E-6) of H = 102 (85–120) km, ηum = 5.1 × 1020 (3.5–7.5)x1020, ηlm = 1.3 × 1022 (0.8–2.8)x1022 where the numbers in parenthesis are 95% confidence limits. The details of the ice sheet, with one exception, are not strongly dependent on the rheological assumptions within this range. The exception is the lower mantle viscosity that remains correlated with the magnitude scaling of the ice sheet: a link that is largely broken by introducing constraints from glacial loading effects on the Earth's rotation and dynamic flattening. The difference between the continental ηum and the comparable estimate of (1–2.5)x1020 for ocean mantle is statistically significant. Shoreline gradient information from Glacial Lakes McConnell, Agassiz, Algonquin and Ojibway provide strong constraints on the response within the interior of the ice sheet and the resulting ice sheet model (LW-6) is characterized by multiple ice domes from at least ∼17–18 ka onwards. The resolution for earlier periods is largely constrained by global ice volume considerations. The two principal domes are over southern Nunavut (the Keewatin Dome) and over Québec-Labrador, both of ≥3500 m thickness, separated by an ice ridge some 1500 m lower than the domes, the latter a requirement imposed by the gradient information. Over western Canada ice thickness gradients east of the Cordilleras are partly constrained by the shoreline data for former Lake McConnell (as is the lateglacial ice thickness of the Keewatin dome) and indicate that any ice-free corridor between the Laurentian and Cordilleran components is unlikely to have existed before ∼13ka. Reconstructions of the glacial lakes are consistent with the locations and timing of the observational evidence for the four major lake systems with the likely drainage routes identified. The evolution of the LW-6 ice-volume function, expressed as equivalent sea level, is characterized by a rapid decrease in ice volume from ∼15 to 14.5 ka, corresponding to the Bølling-Allerød period, in the main from rapid ice retreat along the southern margin, with further contributions from drainage through the St Lawrence River valley and from ice retreat from the major northern straits and gulfs, but not Hudson Strait where the rsl data point to late removal of ice (after 10 ka). The contribution of the drainage of the lakes to global sea level rise is small, < 1 m at the end of 9 ka, with final drainage of the Agassiz, Algonquin and Ojibway lakes occurring through Hudson Strait. The comparison of model predictions based on the above model parameters are, with one exception, consistent with observed GPS rates but do indicate a differential offset between the origins of the geodetic and geophysical reference frames, with the geodetic frame moving relative to the geophysical frame in direction approximately parallel to the earth's rotation axisThis research has at different times been supported by the Australian National University; the chaires internationals de recherche Blaise Pascal financees par la Region Ile-de-France; the Ecole Normale Superieure, Paris; the International Balzan Price Foundation; and the Australian Research Council