Reconcile Mantle Viscosity Structure Inferred from Using ANU and ICE6G Ice Models and Multiple Relative Sea Level Datasets

Two global ice models which describe the spatial and temporal variations of the ice thickness over the last deglaciation have been constructed by different research groups: ICE6G model (Peltier et al., 2015) and ANU model (Lambeck et al., 2010, 2014, 2017). These two ice models are constructed using different datasets of relative sea level (RSL) change observations and different assumptions on mantle viscosity structure. ICE6G is associated with viscosity model VM5a using four viscosity layers with moderate increase in viscosity with depth, while ANU ice model uses two viscosity layers with significantly large increase in the viscosity from the upper mantle to the lower mantle. While the ANU model employs both near field and far field RSL data, ICE6G uses mostly near field RSL data that include different sites from that in ANU model. In this study we seek to explore how these two ice models work with the combined, global RSL data including both near and far field observations and their implications for mantle viscosity. In our study, mantle viscosity is restricted to have two layers, and a large number of forward modeling calculations with different upper and lower mantle viscosities (i.e., ηum and ηlm) for ICE6G and ANU ice models are done to determine the misfits to different sets of RSL data.

Our preliminary results can be summarized as follows. The first is for results using the ANU ice model. The near-field RSL data (i.e., North America and Fennoscandia) are best explained with ηum of ~4.7×1020 Pa s and ηlm of ~2.3×1022 Pa s, consistent with Lambeck et al. (2017). However, the far-field RSL data can be explained by three different viscosity structures for which two are similar to Lambeck et al., (2014), and the third model has ηum of ~1021 Pa s and ηlm of ~1022 Pa s. While both near-field and far-field RSL data are used, the preferred viscosity model is very close to that with only the near-field data. When ICE6G is used, the near-field RSL data prefer ηum of ~1021 Pa s and ηlm of ~5×1022 Pa s, which nearly double the viscosity with the ANU ice model. Viscosity model VM5a with four layers fits similarly well to the RSL data from Peltier et al. (2015) as does our two layers model. However, ICE6G does not fit the far-field RSL data as well as the ANU ice model. We also explore the effects of other possible ice models on fitting the RSL data and viscosity structure.