Towards improving short-term sea ice predictability using deformation observations

Abstract

Short-term sea ice predictability is challenging despite recent advancements in sea ice modelling and new observations of sea ice deformation that capture small-scale features (open leads and ridges) at the kilometre scale. A new method for assimilation of satellite-derived sea ice deformation into numerical sea ice models is presented. Ice deformation provided by the Copernicus Marine Service is computed from sea ice drift derived from synthetic aperture radar at a high spatio-temporal resolution. We show that high values of ice deformation can be interpreted as reduced ice concentration or increased ice damage – i.e. scalar variables responsible for ice strength in brittle or visco-plastic sea ice dynamical models. This method is tested as a proof of concept with the neXt-generation Sea Ice Model (neXtSIM), where the assimilation scheme uses a data insertion approach and forecasting with one member. We obtain statistics of assimilation impact over a long test period with many realisations starting from different initial times. Assimilation and forecasting experiments are run on synthetic and real observations in January 2021 and show increased accuracy of deformation prediction for the first 3–4 d. Similar conclusions are obtained using both brittle and visco-plastic rheologies implemented in neXtSIM. Thus, the forecasts improve due to the update of sea ice mechanical properties rather than the exact rheological formulation. It is demonstrated that the assimilated information can be extrapolated in space – gaps in spatially discontinuous satellite observations of deformation are filled with a realistic pattern of ice cracks, confirmed by later satellite observations. The limitations and usefulness of the proposed assimilation approach are discussed in a context of ensemble forecasts. Pathways to estimate intrinsic predictability of sea ice deformation are proposed.