Improving Arctic sea ice seasonal outlook by ensemble prediction using an ice-ocean model

Abstract

An ensemble based Sea Ice Seasonal Prediction System (SISPS) is configured towards operationally predicting the Arctic summer sea ice conditions. SISPS runs as a pan-Arctic sea ice-ocean coupled model based on Massachusetts Institute of Technology general circulation model (MITgcm). A 4-month hindcast is carried out by SISPS starting from May 25, 2016. The sea ice-ocean initial fields for each ensemble member are from corresponding restart files from an ensemble data assimilation system that assimilates near-real-time Special Sensor Microwave Imager Sounder (SSMIS) sea ice concentration, Soil Moisture and Ocean Salinity (SMOS) and CryoSat-2 ice thickness. An ensemble of 11 time lagged operational atmospheric forcing from the National Center for Environmental Prediction (NCEP) climate forecast system model version 2 (CFSv2) is used to drive the ice-ocean model. Comparing with the satellite based sea ice observations and reanalysis data, the SISPS prediction shows good agreement in the evolution of sea ice extent and thickness, and performs much better than the CFSv2 operational sea ice prediction. This can be largely attributed to the initial conditions that we used in assimilating the SMOS and CryoSat-2 sea ice thickness data, thereafter reduces the initial model bias in the basin wide sea ice thickness, while in CFSv2 there is no sea ice thickness assimilation. Furthermore, comparisons with sea ice predictions driven by deterministic forcings demonstrate the importance of employing an ensemble approach to capture the large prediction uncertainty in Arctic summer. The sensitivity experiments also show that the sea ice thickness initialization that has a long-term memory plays a more important role than sea ice concentration and sea ice extent initialization on seasonal sea ice prediction. This study shows a good potential to implement Arctic sea ice seasonal prediction using the current configuration of ensemble system.