Abstract. Accurate glacial isostatic adjustment (GIA) modelling in the cryosphere is required for interpreting satellite, geophysical and geological records
and for assessing the feedbacks of Earth deformation and sea-level change on marine ice-sheet grounding lines. GIA modelling in areas of active ice loss
in West Antarctica is particularly challenging because the ice is underlain by laterally varying mantle viscosities that are up to several orders of
magnitude lower than the global average, leading to a faster and more localised response of the solid Earth to ongoing and future ice-sheet retreat
and necessitating GIA models that incorporate 3-D viscoelastic Earth structure. Improvements to GIA models allow for computation of the viscoelastic
response of the Earth to surface ice loading at sub-kilometre resolution, and ice-sheet models and observational products now provide the inputs to
GIA models at comparably unprecedented detail. However, the resolution required to accurately capture GIA in models remains poorly understood, and
high-resolution calculations come at heavy computational expense. We adopt a 3-D GIA model with a range of Earth structure models based on recent
seismic tomography and geodetic data to perform a comprehensive analysis of the influence of grid resolution on predictions of GIA in the Amundsen
Sea Embayment (ASE) in West Antarctica. Through idealised sensitivity testing down to sub-kilometre resolution with spatially isolated ice loading
changes, we find that a grid resolution of ∼ 13 of the radius of the load or higher is required to accurately capture the elastic
response of the Earth. However, when we consider more realistic, spatially coherent ice loss scenarios based on modern observational records and
future ice-sheet model projections and adopt a viscoelastic Earth, we find that predicted deformation and sea-level change along the grounding line
converge to within 5 % with grid resolutions of 7.5 km or higher, and to within 2 % for grid resolutions of 3.75 km and
higher, even when the input ice model is on a 1 km grid. Furthermore, we show that low mantle viscosities beneath the ASE lead to viscous
deformation that contributes to the instrumental record on decadal timescales and equals or dominates over elastic effects by the end of the 21st
century. Our findings suggest that for the range of resolutions of 1.9–15 km that we considered, the error due to adopting a coarser grid
in this region is negligible compared to the effect of neglecting viscous effects and the uncertainty in the adopted mantle viscosity structure.