Constraints of Relative Sea Level Change on the Late Pleistocene Deglaciation History

In this study, we examine the relationships among mantle viscosity, ice models and RSL data. We analyzed two widely used ice models, the ANU and ICE-6G ice models, and found significant difference between these two models, suggesting that significant uncertainties exist in ice models. For six RSL datasets covered both the near- and far-field from published works [Peltier et al., 2015; Lambeck et al., 2014, 2017; Vacchi et al., 2018; Engelhart et al., 2012, 2015], we performed forward GIA modelling using a 1-D compressible Earth model to seek the preferred upper and lower mantle viscosities that fit each of the six RSL datasets, for each of these two ice models. Our calculations show that viscosity in the lower mantle is significantly larger than the upper mantle for almost all the pairs of RSL datasets and ice models, but the RSL datasets for North America and Fennoscandia by Peltier et al., [2015] can be matched similarly well with a large parameter space of upper and lower mantle viscosities, both relatively uniform mantle viscosity and with large increase with depth. The preferred mantle viscosity using the ANU ice model and Lambeck et al. [2017] RSL data for North America is in a good agreement with that by Lambeck et al. [2017].    By using the GIA model with the preferred viscosity structures, we constructed the spatial and temporal distributions of misfit to different RSL datasets, for both the ICE-6G and ANU ice models. The misfit patterns for the ANU and ICE-6G ice models do not differ significantly in North America, although these two ice models differ greatly in North America. However, due to relatively small ice volume in ICE-6G, it fails to explain the far-field RSL data, reflecting the so-called “missing ice” problem. Guided by the spatial and temporal misfit patterns, we made initial attempts to modify ICE-6G by adding more ice to the ice model to improve the fit to far-field RSL data. The three modified ICE-6G ice models we consider all significantly improve far-field RSL data, while maintaining or even improving misfit for near field RSL data. This shows the promise with our method in improving ice models and fit to RSL data.