TOPAZ4: an ocean-sea ice data assimilation system for the North Atlantic and Arctic

• Published in: Ocean Science
• Main Authors: Sakov, P., Counillon, F., Bertino, L., Lisæter, K. A., Oke, P. R., Korablev, A.
• Format: Text
• Language: English
• Published: 2018
• Subjects: Arctic; Arctic Ocean; Fram Strait; North Atlantic; Sea ice
• Online Access: https://doi.org/10.5194/os-8-633-2012; https://os.copernicus.org/articles/8/633/2012/

We present a detailed description of TOPAZ4, the latest version of TOPAZ – a coupled ocean-sea ice data assimilation system for the North Atlantic Ocean and Arctic. It is the only operational, large-scale ocean data assimilation system that uses the ensemble Kalman filter. This means that TOPAZ features a time-evolving, state-dependent estimate of the state error covariance. Based on results from the pilot MyOcean reanalysis for 2003–2008, we demonstrate that TOPAZ4 produces a realistic estimate of the ocean circulation in the North Atlantic and the sea-ice variability in the Arctic. We find that the ensemble spread for temperature and sea-level remains fairly constant throughout the reanalysis demonstrating that the data assimilation system is robust to ensemble collapse. Moreover, the ensemble spread for ice concentration is well correlated with the actual errors. This indicates that the ensemble statistics provide reliable state-dependent error estimates – a feature that is unique to ensemble-based data assimilation systems. We demonstrate that the quality of the reanalysis changes when different sea surface temperature products are assimilated, or when in-situ profiles below the ice in the Arctic Ocean are assimilated. We find that data assimilation improves the match to independent observations compared to a free model. Improvements are particularly noticeable for ice thickness, salinity in the Arctic, and temperature in the Fram Strait, but not for transport estimates or underwater temperature. At the same time, the pilot reanalysis has revealed several flaws in the system that have degraded its performance. Finally, we show that a simple bias estimation scheme can effectively detect the seasonal or constant bias in temperature and sea-level.