What’s the difference between MPAS-DART and MPAS-JEDI?

[ZIZU296]

Hi everyone,
I’m working with the MPAS model and recently started looking into its data assimilation systems. I found that there are two frameworks, MPAS-DART and MPAS-JEDI, but I’m a bit confused about how they differ in terms of design, applications, and development status. Could someone explain the main differences between them and which one is more suitable for current or future MPAS work? Any clarification would be really helpful. Thanks a lot!

 

[Ming Chen]

My understanding of MPAS-DART and MPAS-JEDI is that, MPAS-JEDI is a data assimilation interface for MPAS, JEDI means the Joint Effort for Data Assimilation Integration (JEDI) system, which uses a variational approach. MPAS-DART is the data assimilation interface for the Data Assimilation and Research Toolkit (DART) for MPAS, which is built on ensemble-based Kalman filter approach.
MPAS-JEDI is being actively developed and widely used. I would suggest that you follow the approach of MPAS-JEDI in your research.

 

[ZIZU296]

My understanding of MPAS-DART and MPAS-JEDI is that, MPAS-JEDI is a data assimilation interface for MPAS, JEDI means the Joint Effort for Data Assimilation Integration (JEDI) system, which uses a variational approach. MPAS-DART is the data assimilation interface for the Data Assimilation and Research Toolkit (DART) for MPAS, which is built on ensemble-based Kalman filter approach.
MPAS-JEDI is being actively developed and widely used. I would suggest that you follow the approach of MPAS-JEDI in your research.

Thank you very much for your clarification. So, I see, the JEDI system is more like finding a single optimal solution through mathematical optimization, while the ensemble-based Kalman filter approach generates many possible states and then takes their average as the best estimate.