On the substantial influence of the treatment of friction at the grounding line

Abstract

The dynamical contribution of marine ice sheets to sea level rise is largely controlled by grounding line (GL) dynamics. Seroussi et al. (2014) emphasised the sensitivity of numerical ice flow model results to the practical implementation of the friction of the ice on its bed in the very close vicinity of the GL. Elmer/Ice is a reference finite element (FE) ice flow model used in recent marine ice sheet model intercomparison (MISMIP) exercises. In the model, the GL is defined as the nodes where the ice is in contact with the bedrock but belong to both grounded and floating elements. Inherently to the FE method, computing the contribution of the friction by element requires evaluating the friction at the integration points. In Elmer/Ice, this is done by interpolating the values of the friction parameter C prescribed at the nodes. In this brief communication, we discuss and compare three alternative ways to prescribe the friction at the GL: (i) C is prescribed and non null at the GL nodes, (ii) C is set to zero at the GL nodes, and (iii) C is discontinuous at the GL nodes (i.e. is prescribed and non null for grounded elements and otherwise null). So far, all published results using Elmer/Ice were obtained with the first method. Using the MISMIP3d diagnostic experiment, we first show that, although the change in the total force at the base is insignificant, the three methods lead to significantly different velocity fields. We then show that these methods also lead to different steady state GL positions and different transient behaviours. Such model sensitivity to the methods discussed here is certainly specific to the high friction prescribed in the MISMIP experiments and should be smaller in real setups where friction in the vicinity of the GL would be expected to be lower. Results obtained with the three methods are available as Supplement for future comparisons.