4DVAR minutes 03/06/06

AFWA 4D-Var meeting Hans, Dale, Qingnong, John M, Tom, John B, Chris 2:00-3:30 pm, 03/06/2006. Lilly room

1. 4D-Var activities in the past two months

1) Dale, John and Hans visited AFWA, 18-20 January. They got the 4D-Var prototype installed and tested on the AFWA computer system. One of the important results from this visit is the test on a 80x80x17 grid, which makes meteorological tests possible.

2) Hans presented the 4D-Var work at AFWA and at the AFRL outbrief at MMM.

3) Hans started testing the 80x80x17 configuration on bluesky and bluevista. For 3 iterations, it took 4.39 hours on bluesky, 3.2 hours on bluevista, but 2.25 on AFWA computer.

4) Dale, John B, Zhiquan and Hans started to get accounts on AFWA computer systems for testing WRF-Var.

5) John M, John B, Wei Huang and Hans had a meeting on merging 4D-Var branch to the main trunk. Some steps were taken to rearrange the Registry.

6) John M has made nup and Hans tested it.

2. The ongoing and the planned work

1) Meteorological tests

Hans will continue the test on bluevista and possibly on AFWA’s computer.

It is time to assess the 4D-Var performance on the meteorological side.

2) Merge the 4D-Var branch

John B started the 4D-Var merging while waiting for Dale’s comments on the alpha control variable merging code. The major problems are in wrfplus. Now even the WRF nonlinear model does not run. Tom explained his work before John B took over. John B outlined his steps as: 3D-Var (radiance, alpha control variable, etc) – WRF model – wrfplus – new WRF.

John B estimated the work be done by 15 April.

Qingnong raised the issue of compiling time. John B is working on the issue. There was also a lengthy discussion on the policy of code merging.

3) Parallel code and coupler

After some explanation and discussion, it was agreed to add one task before these two. The order of the work should be: job control – parallel – coupler. John B estimated the starting time of these tasks to be mid April. John M will join the work from that point. He estimated 4 months would be needed to complete the work. John B asked the deliver time for the basic system. Dale confirmed that the basic system should, as other AFWA deliverables, be delivered in February 2007. Other 4D-Var deadlines are internal in the group.

4) Nup inverse

Hans mentioned needs for nup inverse, which puts the low-resolution analysis increments to high resolution. John M wrote an email earlier asking for a choice among solutions for lateral boundaries. He explained the problem again at the meeting. Dale commented that accuracy of nup is much more important than its inverse as the lateral boundary treatment in WRF-Var has several known errors. John B suggested giving a low priority to this work. Hans still wished to have it early so the full multi-incremental formulation can be coded and tested.

Hans and Dale will give John M a more precise description of the nup inverse.

5) Physics adjoints Qingnong mentioned the importance of diffusion in hurricane simulations. The diffusion adjoint is an AFWA task. He will start the work soon. In the same time, through his Mississippi project, he will invite a visitor to work on the NCAR microphysics package. Qingnong also reminded the group the TAF license and suggested using TAF to generate more physics adjoints in this year. Hans talked with Laura recently. Laura is working on an implementation of the CSU nonlinear microphysics package and plans to start the TL and AD work after summer with us.

3. AOB

• Hans is working with Xin Zhang (Hawaii Univ.) on a proposal to implement JcDFI. • Hans will write a memo and distribute it to the group for comments. • The next meeting will be around 15 April after the 4D-Var merge.