4D-Var meeting

09/28/2007

2pm, 2072

Participants: Dale, Chris, Michael, Jenny, Xiaoyan, Qingnong, Xin, Tom, Yong-Run, Hans, (have I forgotten someone?)

1. Transition of the 4D-Var lead

This is the transition meeting for the 4D-Var lead. Dale will take over the lead after the meeting.

Hans will start at NASA on Jan 2. In the remaining 3 months, he will 1) finish the 4D-Var manuscript with Dale and Yongsheng; 2) start the tech note or manual for 4D-Var; and 3) help Dale with the all 4D-Var management related issues.

2. September report

Xin:

(work at NRL on another project)

Investigated the solve_em codes and began to write and modify codes to reduce duplicate computation in adjoint integration. No clear results yet.

Tom:

Tracked down cause of bitwise differences between runs of the Tangent Linear model made using different numbers of MPI tasks. Fixed bug by adding new halo update "HALO_EM_SD_TL" before call to g_surface_drag(). Tested using Xin Zhang's TL test case. Repaired code now gets bitwise identical results for 1, 4, and 8 MPI tasks.

Yongsheng:

1) produce forecasts and obs-space verification on 45-km Haitang case

2) making forecasts for 45-15km multi-incremental Haitang case experiments

3) attempting sensitivity study for hurricane forecast

Hans:

Data assimilation experiments, Haitang case (3 days):

1) 15km 3D-Var cold-start

2) 15km 3D-Var cycling

3) 15/45 km 4D-Var REF (centered) cold-start

4) 15/45 km 4D-Var REF (centered) cycling

Qingnong:

Continue to debug the TL and AD code of the cumulus scheme.

3. Overview of the AFWA 4D-Var tasks

Hans went through the AFWA 4D-Var tasks.

In summary, we are slightly behind the original schedule (internally set) but should have no problem to finish most of the tasks on time.

Tom spent weeks to fix bugs in the TL and AD parallelization code. There are still some halo-update related subroutines to be optimized, but he may have no time on this due to the WRF 3.0 release plan. He and Dale will check with Kathy how much time he has charged on his 4D-Var task.

We cannot do J_lbc task as both Hans and Xin change jobs soon.

Dale also wants to cancel the A2C task and will contact Steve Rugg about the task cancellations.

Qingnong wish to increase the budget for cumulus task, from 0.4 to 0.5.

4. AOB

Hans thanked for all the support he has received in the past 3 years. He suggested having bug-fix related tasks in the future work plans.

Dale thought the 4D-Var development is on schedule and we are now entering application era. There are at least 5 proposals based on 4D-Var. He hoped Hans could spend the next 3 months on the paper and tech note.

Dale planned to have a strategy meeting to discuss:

1) 4D-Var code managing

2) OpenAD

3) Other TL and AD code: MM5 or Perturbation Model

4) Leapfrog in wrfplus

5) Long-term plan

Hans agreed to give a seminar and/or tutorial on 4D-Var.

Dale suggested having bi-weekly 4D-Var meetings. All agreed. Next meeting: Tuesday 16 Oct, room 2072.