Imaging 3D mantle viscosity beneath Greenland using relative sea level and GNSS observations with an adjoint approach: A test of discordant GIA uplift models.

Abstract:

The recent development of the adjoint method within the context of Glacial Isostatic Adjustment (GIA) provides a unique opportunity to perform data driven inversions of key GIA parameters, such as mantle viscosity. Central to this approach is the accurate and efficient determination of sensitivity kernels for a range of possible GIA observables (e.g., relative sea level, RSL, measurements and GNSS deformation rates). These kernels require just two numerical simulations – a forward simulation driven by the ice history and an adjoint simulation driven by a ‘fictitious’ load applied at the observation point. This ‘fictitious’ load when weighted by an appropriate misfit functional results in the gradient of the misfit function with respect to the model parameter(s), which can be used in a gradient based optimization scheme to iteratively update the model parameter(s) and minimize the data misfit. Here we apply this approach, assuming Maxwell viscoelasticity, to image the 3D mantle viscosity structure beneath Greenland using near-field relative sea level measurements since the Last Glacial Maximum and vertical GIA uplift rates determined by GNSS observations (Schumacher et al., 2018). Because these data sets suggest markedly different GIA uplift rates for Greenland (Kahn et al., 2016), we leverage adjoint inversions of 3D mantle viscosity as a diagnostic tool to investigate whether both data sets can be simultaneously fit equally well and whether similar 3D mantle viscosity structures are imaged. To do this, we perform the inversion for each data set independently, as well as jointly. We note that the resulting viscosity models are non-unique and will be affected by both the ice history and the starting viscosity model. To explore these effects, we test different ice histories and different starting viscosity models. Finally, the imaged viscosity structures will be compared to independent estimates of mantle viscosity based on seismic tomography and rheological laboratory experiments.