Data Assimilation

Data assimilation is a method used in numerical modeling to assure that model results are close to observations. By using a data assimilation system, we can obtain “analysis fields”, which are a set of ocean and sea ice variables optimally combining observations and a model. The analysis fields are applied for a variety of ocean and sea ice studies in the Arctic Ocean, and also provide initial conditions for future projection experiments (e.g., sea ice outlook).We have been developed data assimilation system which exploits observed sea ice concentration, ice thickness and sea ice drift obtained from satellite measurements. A 4-dimensional variational (4DVar) data assimilation system and a genetic algorithm optimization system are currently used to synthesize model results and observed data in the Arctic Ocean. The systems have been successfully applied to North Atlantic/Arctic Ocean Sea Ice Model (NAOSIM) developed at the Alfred Wegener Institute (AWI). We are currently working to apply the systems to Finite-volumE Sea ice-Ocean Model (FESOM2), also developed in AWI.

Contact:

Dr. Hiroshi Sumata

Seasonal cycle (left) and interannual variation of sea ice extent (more than 15% sea ice cover). Black: observation, blue: model simulation without data assimilation, red: model simulation with data assimilation. (Graphic: Alfred-Wegener-Institut)
Spatial pattern of sea ice extent in September, from 1990 to 2010 (every 5 years). Black: observation, blue: model simulation without data assimilation, black: model simulation with data assimilation. (Graphic: Alfred-Wegener-Institut)
Spatial pattern of sea ice motion in August 2006: Left panel: model simulation without data assimilation, center panel: model simulation with data assimilation, right panel: observation. (Photo: Alfred-Wegener-Institut)