Lateral heterogeneities in mantle viscosity and post‐glacial rebound

Seismic tomographic results and convection calculations support the evidence of horizontal temperature variations in the mantle. On the basis of commonly accepted rheological laws, we thus expect lateral variations in the viscosity of several orders of magnitude. This paper is concerned with the effects of lateral viscosity variations on vertical displacement induced by Pleistocenic deglaciation. A finite‐element scheme in axial symmetry mimics the relaxation process of a flat Earth model characterized by a linear Maxwell rheology. We follow a spectral approach to analyse the impact of lateral viscosity variations with different amplitudes and wavelengths. The potential impact of lithospheric thickening and viscosity increase in stable continental regions on the interpretation of sea level data is also analysed. Lateral viscosity heterogeneities are found to have strong influence on ground deformations induced in deglaciated areas. From the analysis of sea level data near the centre of the former Fennoscandian ice sheet and the theoretical predictions of radially and laterally stratified Earth models, we explore the extent to which lateral viscosity contrasts may have influenced the inference of long‐term mantle viscosity. While radially stratified models predict a rather uniform mantle viscosity, we show that lateral viscosity contrasts of 1 or 2 orders of magnitude are consistent with sea level data if the wavelength of the heterogeneity is comparable with the dimension of a convecting cell. In this case, the average viscosity can deviate by around an order of magnitude from the ‘canonical’ value of 1021 Pa s predicted by rebound models in the past. Viscosity values close to 1021 Pa s are found to be appropriate for the mantle region underlying the load. Long wavelength viscosity variations of 3 or 4 orders of magnitude degrade our ability to reproduce the observed uplift in the centre of the ice sheet and must be ruled out. For wavelengths comparable with the horizontal extension of the surface load and viscosity contrasts of 1 or 2 orders of magnitude, model results are found to be inconsistent with sea level data. These findings suggest that rebound modelling can become a useful tool to constrain the magnitude and wavelength of viscosity contrasts. Comparison between horizontally and vertically varying models indicates that lateral viscosity heterogeneities could have been interpreted as radial variations in previous rebound studies. Copyright © 1989, Wiley Blackwell. All rights reserved