Abstract. Poor convergence with resolution of ice sheet models when simulating
grounding line migration has been known about for over a decade. However,
some of the associated numerical artefacts
remain absent from the published literature. In the current study we apply a Stokes-flow finite-element marine
ice sheet model to idealised grounding line evolution experiments.
We show that with insufficiently fine model
resolution, a region containing multiple steady-state grounding
line positions exists, with one steady state per node of the model
mesh. This has important implications for the design
of perturbation experiments used to test convergence of grounding
line behaviour with resolution.
Specifically, the design of perturbation experiments can be under-constrained,
potentially leading to a “false positive” result.
In this context a false positive is an experiment that appears
to achieve convergence when in fact the model configuration is
not close to its converged state.
We demonstrate a false positive: an apparently successful
perturbation experiment (i.e. reversibility is shown) for a
model configuration that is not close to a converged solution.
If perturbation experiments are to be used in the future,
experiment design should be modified to provide additional
constraints to the initialisation and spin-up requirements. This region of multiple locally stable steady-state
grounding line positions has previously
been mistakenly described as neutral equilibrium.
This distinction has important implications for understanding
the impacts of discretising a forcing feedback involving
grounding line position and basal friction.
This forcing feedback cannot, in general, exist in a region of
neutral equilibrium and could be the main cause of poor convergence
in grounding line modelling.