Proposed Design of the Third Marine Ice Sheet Model Intercomparison Exercise

Thursday, 18 December 2014: 9:30 AM
Stephen L. Cornford1, Gudmundur Hilmar Gudmundsson2, Frank Pattyn3, Gaël Durand4, Daniel F Martin5 and Antony John Payne1, (1)University of Bristol, Bristol, United Kingdom, (2)NERC British Antarctic Survey, Cambridge, CB3, United Kingdom, (3)Universite Libre de Bruxelles, Brussels, Belgium, (4)LGGE Laboratoire de Glaciologie et Géophysique de l’Environnement, Saint Martin d'Hères, France, (5)Lawrence Berkeley National Laboratory, Berkeley, CA, United States
The MISMIP and MISMIP3D marine ice sheet model intercomparison exercises have become popular benchmarks, and several modeling groups have used them to show how their models compare to both analytical results and other models. Both sets of experiments studied grounding line migration in response to changes to (in effect) properties of the grounded ice, either perturbations to the basal traction coefficient (C), or to ice softness (A). In particular, both exercises concluded that careful treatment of the grounding line was required in numerical models, either through fine resolution (typically better than 1 km), or higher order methods, or modifications to the discretization scheme.

Given the success of these experiments, we propose a sequel. Like MISMIP3D, the experiments take place in a idealized, three-dimensional setting and compare full 3D (Stokes) and reduced, hydrostatic flow models. Unlike the earlier exercises, the primary focus will be the response of models to changes in ice shelf conditions, especially sub-shelf melting. The chosen configuration features an ice shelf that experiences substantial lateral shear and buttresses the upstream ice, and so is well suited to melting experiments. We also aim to design an experiment in which differences between the steady states of each model are minor compared to the response to melt-rate perturbations, a choice that reflects typical real-world applications where parameters are chosen so that the initial states of all models tend to match observations (and hence each other).

We will present the MISMIP+ experiments, and show how results computed with the BISICLES ice sheet model are affected by mesh resolution, choice of flow model, and choice of melt-rate.