Examples: Adjoint Sensitivities and Assimilation in the Arctic

The Arctic region is a recent application area for adjoint techniques. We are working on this turf together with OASys within the project DAMOCLES. Our activities build on the adjoint version (ADNAOSIM) of the coupled ocean-sea ice model NAOSIM that we generated by TAF. NAOSIM itself was developed in Rüdiger Gerdes's group at the Alfred Wegener Institute. In parallel, the colleagues at the ECCO consortium are applying the TAF-generated adjoint of their ocean general circulation model to study the Arctic.

One of ADNAOSIM's first applications is an analysis of the 2007 sea ice minimum, a dramatic event that received widespread media attention. In the study we use the adjoint to trace back the sensitivities of the total ice covered area in early September 2007 with respect to the model's surface boundary and initial conditions The corresponding publication can be found here. You can watch a set of animations that show the sensitivities with respect to wind stress and 2 m temperature going backwards from September to March. You can also watch a set of animations that show the the sensitivities with respect to the inital temperature and salinity, going from the uppermost to the lowest model layer.

ADNAOSIM is also the core of the variational assimilation system NAOSIMDAS, built around NAOSIM. One recent application of NAOSIMDAS is the production of simulated model fields that combine the information from NAOSIM with a range of observational constraints. First results are described on this poster. Another NAOSIMDAS application is the Sea Ice Outlook, a kind of competition within the Arctic science community to predict the minimum sea ice extent for the upcoming September. These predictions start with the beginning of the melting season in June and are updated in monthly intervals. Details on the 2009 Sea Ice Outlook series can be found in this report .

Copyright © FastOpt - all rights reserved