CISL Seminar- Strongly Coupled Physical and Biogeochemical Data Assimilation: Present and Future

12/09/2015 - 9:30am
Mesa Lab - Main Seminar Room

Hajoon Song

Massachusetts Institute of Technology

Biogeochemical (BGC) state estimation is in high demand as new autonomous observation platforms have extended the observational space more than ever. Conventional data assimilation methods assuming Gaussian error distribution, however, fail to estimate BGC fields accurately because of their skewed error distributions and positive-definite proper-ties. In this talk, I will present a new log-normal 4-dimensional variational approach for BGC data assimilation, a strongly coupled physical and BGC assimilation system, and its implementation in the regional ocean modeling system (ROMS).

The log-transformed approach for BGC data assimilation outperforms the traditional approach with Gaussian error statistics and implicitly ensures the positive definite property, making this approach attractive for BGC variables. It is then fully coupled to the existing physical data assimilation. This hybrid data assimilation system allows BGC observations to directly influence physical model fields while allowing physical observations to drive BGC fields. In an application to the California Current System for the year 2000, the coupled data assimilation system successfully fits eddy-resolving ROMS to the surface chlorophyll and physical observations, and reduces the root-mean-square error of the surface chlorophyll by 40 percent. This approach has broader implications because lognormal error distributions well represent other variables such as precipitable water and sea-ice in the climate system. For future development, I will discuss how this coupled data assimilation system can be further improved by ensemble data assimilation methods and how ensemble methods can include coupled dynamics between physical and BGC components.

Date: Wednesday December 9, 2015

Time: 9:30am (this seminar will start 30 minutes earlier than usual)

Location: NCAR- Mesa Laboratory- Main Auditorium