Strongly Coupled Data Assimilation and Initialization with the Parallel Data Assimilation Framework (PDAF)

• Main Authors: Nerger, Lars, Tang, Qi, Mu, Longjiang
• Format: Conference Object
• Language: unknown
• Published: 2021
• Subjects: Arctic
• Online Access: https://epic.awi.de/id/eprint/55045/; https://epic.awi.de/id/eprint/55045/1/poster_AWICM-PDAF.pdf; https://hdl.handle.net/10013/epic.df856e03-b725-44d1-b948-1959fa559a48

The Parallel Data Assimilation Framework (PDAF, http://pdaf.awi.de) is an open-source software framework for highly efficient ensemble data assimilation with complex models on supercomputers. PDAF was developed to simplify the generation of a data assimilation system from existing models. For coupled data assimilation, PDAF is used for example with the coupled atmosphere-ocean model AWI-CM, with different coupled ocean biogeochemical models, and with the atmosphere-land surface-subsurface model TerrSysMP. However, there is a wide range of further applications of PDAF. PDAF provides functionality to perform ensemble integrations, which can be used for ensemble predictions and ensemble data assimilation. Further, PDAF provides several fully-implemented ensemble filter and smoother methods for data assimilation. One can build the data assimilation application either by using model restart files or by directly augmenting the different compartment models of a coupled system with data assimilation functionality. The ensemble data assimilation can then be applied in an efficient way with complex models like AWI-CM on supercomputers with excellent scalability and efficiency. PDAF directly supports both weakly and strongly coupled data assimilation. Discussed will be the features of PDAF and the structure of data assimilation systems for coupled data assimilation with PDAF.