A data-calibrated distribution of deglacial chronologies for the North American ice complex from glaciological modeling

Past deglacial ice sheet reconstructions have generally relied upon discipline-specific constraints with no attention given to the determination of objective confidence intervals. Reconstructions based on geophysical inversion of relative sea level (RSL) data have the advantage of large sets of proxy data but lack ice-mechanical constraints. Conversely, reconstructions based on dynamical ice sheet models are glaciologically self-consistent, but depend on poorly constrained climate forcings and sub-glacial processes. As an example of a much better constrained methodology that computes explicit error bars, we present a distribution of high-resolution glaciologically-self-consistent deglacial histories for the North American ice complex calibrated against a large set of RSL, marine limit, and geodetic data. The history is derived from ensemble-based analyses using the 3D MUN glacial systems model and a high-resolution ice-margin chronology derived from geological and geomorphological observations. Isostatic response is computed with the VM5a viscosity structure. Bayesian calibration of the model is carried out using Markov Chain Monte Carlo methods in combination with artificial neural networks trained to the model results. The calibration provides a posterior distribution for model parameters (and thereby modeled glacial histories) given the observational data sets that takes data uncertainty into account. Final ensemble results also account for fits between computed and observed strandlines and marine limits. Given the model (including choice of calibration parameters), input and constraint data sets, and VM5a earth rheology, we find the North American contribution to mwp1a was likely between 9.4 and 13,2 m eustatic over a 500 year interval. This is more than half of the total 16 to 26 m meltwater pulse over 500 to 700 years (with lower values being more probable) indicated by the Barbados coral record (Fairbanks, 1989; Peltier and Fairbanks, 2006) if one assumes a 5 meter living range for the Acropora Palmata coral. 20 ka ice volume for North America was likely 70.1 +/- 2.0 m eustatic, or about 60% of the total contribution to eustatic sea level change. We suspect that the potentially most critical unquantified uncertainties in our analyses are those related to model structure (especially climate forcing), deglacial ice margin chronology, and earth rheology. (C) 2011 Elsevier B.V. All rights reserved.