hide
Free keywords:
-
Abstract:
Accurate assessment of past and future ice sheet-driven sea level variations requires continental-scale
numerical models that are able to cover long response time scales of ice sheets, typically involving
thousands of years. In order to allow such time scales, the Shallow Ice and Shallow Shelf approximations
(SIA/SSA) are commonly introduced to simplify the Stokes equations describing the dynamical
regimes of the grounded and
oating ice sectors, respectively. However, the SIA is inapplicable to
the fast
owing regions where basal sliding operates and some of the neglected terms in the Stokes
equations become important. Since these terms are included in the SSA, many recent studies
have introduced heuristic \hybrid" combinations of both approximations in order to reproduce the
dynamics of rapid ice
ow zones. However, a side-by-side evaluation of their performance in realistic
scenarios is still missing. In this study, we implement three dierent hybrid approaches into the same
continental-scale numerical model of the Antarctic Ice Sheet. Our experiments involve an automated
calibration of the model based on an iterative technique that uses modern observational data sets
to estimate variations in the poorly constrained basal sliding parameters. For validation purposes,
we compare the results with the independent observed surface velocity eld. Our results show that
the automated calibration compensates for the dierences in the hybrid schemes through dierent
parameter distributions, thereby producing similar ice sheet congurations. We use an averaged parameter
distribution to demonstrate the in
uences arising from the use of dierent hybrid approaches.
Although the uncertainty in the basal conditions limits an objective evaluation of the hybrid schemes,
our experiments show that the retrieved parameter distributions are not exchangeable between the
schemes. This suggests that the results of the calibration and/or initialization of a specic numerical
ice sheet model cannot be straightforwardly transferable to a model that uses a dierent level of
approximation of the Stokes equations, which represents an important limitation for more complex
and computationally expensive ice sheet models.
