Real-time estimation of Arctic sea ice thickness through maximum covariance analysis

摘要: A challenge for model-based seasonal predictions of sea ice is an accurate representation initial conditions, particularly sparsely observed thickness (SIT). The Canadian Seasonal to Interannual Prediction System (CanSIPS) currently initializes SIT by nudging simulated values toward a climatology. To improve on this, we use data from Pan-Arctic Ice Ocean Modeling and Assimilation investigate how accurately can be estimated in real time using better physically relevant predictors. We (1) test the skill several predictors maximum covariance analysis (MCA), (2) apply approach which blends concentration lagged (4 month averaged) level pressure, (3) compare this method against current CanSIPS initialization scheme over 1981–2012. MCA-based statistical model reduces areal mean temporal absolute errors 48% relative shows consistent estimating volume all months (r = 0.95).