Challenging model-based forecasts of the sea-ice edge with a spatial probabilistic approach to damped anomaly persistence

Abstract:

Recent advancement in dynamical sea-ice models have enabled weather agencies to forecast sea-ice conditions at sub-seasonal to seasonal timescales. In previous studies using the S2S dataset, the ice-edge output of various forecasting centers was compared against reference forecasts to assess the predictive skill of the models. However, the simplest types of reference forecasts – persistence of the initial state and climatology – do not exploit the observations optimally and thus lead to artificially high forecast skill. For spatial objects such as the ice-edge location, the development of damped-persistence forecasts that combine persistence and climatology in a meaningful way poses a challenge. With this motivation, we have developed a probabilistic reference forecast method based purely on ice-edge observations that combines the climatologically derived probability of ice presence with initial anomalies of the ice edge. We have tested and optimized the method based on minimization of the Spatial Probability Score (SPS), and compared it to the output from other models in the S2S dataset. Besides SPS we also applied the Modified Hausdorff Distance (MHD) as verification metric to make sure that the results are robust to the choice of the metric. We find that the resulting reference forecasts provide a challenging benchmark to assess the added value of dynamical forecast systems.