Progress review

Bug fix (Xin and Hans)

Xin identified the bug in boundary input in WRF TL and AD. The boundary input codes in WRF+ were designed based on the boundary condition generated by WRF REAL 2.0, which has a diifferent name with the current WRF REAL 2.2 generated BC. Then, the BC can not be read in both WRF TL and AD.

During his one-week visit (talk to Qingnong, Jimy and others), Xin also identfied another bug in Registry related codes which almost totally removed the BC input codes.

Both bugs were fixed. Hans started another round of tests with the new code.

Physics (Qingnong, Jimy, Xiaoyan)

Still work on the adjoint code testing.

Meteorological testing (Yongsheng and Hans)

Reruns of the Haitang case have been evaluated. The global bogusing data lead to improvement in cold-start runs. For the cycling experiments, the one-sided 4dvar seems to give better scores than the centered window 4dvar.

Another round of tests with the new code has started.

Multi-incremental (Hans)

The data assimilation experiments using the multi-incremental 4dvar was postponed due to the bug fixes and the additional experiments after that.

Optimization (Tom)

Tom will start to work on the parallel TL code. Currently parallel code cannot reproduce the serial run results.

The work on the halo update optimization will be resumed after the bug fix.

Other tasks

Dale suggested to include 4dvar in the regression tests.

We also had a short discussion on the 4dvar paper writing and a tech note (be part of WRF-Var note or a separate note).

AOB

The next meeting will be 10:00 September 28th, room 2072.