Nudging an inversion-based ice sheet model towards observed mass change during model relaxation: application to the C.H. Ostenfeld Glacier

Abstract. Snapshot inversions are widely used to initialize numerical ice sheet models, but they do not necessarily produce a dynamically consistent state for transient simulations. The transition from an inverted snapshot to a prognostic simulation can generate a substantial initial model shock in ice thickness, velocity, and potentially unrealistic transitions between grounded and floating ice. To reduce these effects, modellers commonly use a relaxation period during which the ice sheet model is allowed to adjust towards a dynamically consistent state, often at the expense of increasing errors relative to the observed present-day geometry and velocity. Here, we introduce a nudging method for that relaxation period that reduces the modelled ice geometry drift while allowing the model to adjust dynamically. The method is designed to (i) reduce model drift, (ii) limit the increase of errors in ice geometry with respect to observations during the relaxation period, and (iii) enable transient simulations to be consistent with observed present day thinning rates. We apply the method to the C.H. Ostenfeld Glacier in northern Greenland. We find that nudging substantially reduces the drift in ice geometry and grounding-line position during the relaxation period and that it enables the model to capture the observed present-day thickness change rate. The method provides a viable, simpler, and cheaper alternative to a transient calibration, commonly used in spin-up based ice sheet models. The method is not ice sheet model-specific and will, when implemenetd widely across models, contribute to reducing the uncertainty in sea level rise projections by better constraining model initial dynamic states.