The meeting was cancelled due to the spring break and reports from Hans, Tom, Xin, Jimy and Qingnong were compiled as the "meeting memo" for March 2007.
Worked together with Xin on testing the new code over the Haitang case, KMA case and JcDF.
Prepared the annual review for AFWA briefing and finalized the 2007-2008 4dvar workplan.
Gave a seminar in Univ of Hawaii: 4-dimensional variational data assimilation for WRF.
Spent some time with Xin on debugging xatowrf_ad|tl and JcDF. The work is still ongoing.
Began working on optimization of 4DVAR adjoint via trade-off of re-computation for memory. Set up a baseline test case with assistance from Xin. Analyzed dependencies in solve_em_ad.F and identified loop "runge_kutta_loor" as low-hanging fruit. Wrote code to replace redundant computation in this loop with storage in memory. This new behavior is switched on and off by a new namelist variable. Developed a very simple buffering toolkit to provide storage and retrieval. Initial cut at code is complete and ready to test. Thomas Nehrkorn and I began discussing how he might contribute to the 4DVAR optimization effort. We are drafting a proposal.
I had further discussions about OpenAD with Chris Hill (MIT) who is using both TAF and OpenAD for the MITgcm. Chris asked if we might know of anyone
(presumably including us) who might want to help support OpenAD development. Do we?
Qingnong and I have reviewed the output from the automatic test generator and discovered one more bug which I have fixed. Work on the automatic test
generator is now on hold while the 4DVAR optimization takes top priority.
Cooperated with Qingnong Xiao and Xiaoyan Zhang, the AD and TL codes of Large scale condensation were merged into wrfplus and both TL and AD check pass.
Modifying nup-down codes toward WRF2.0 compatible.
Fixed bugs in AD&TL codes of transfer between wrfvar space and wrfplus space.
Continue development of a simplified cumulus parameterization as part of the WRF-Var simplified physics suite to use in adjoint code. To date, an algorithm has been designed, but coding is not yet done.
Worked with Xin on debugging the large-scale condensations code in 4DVAR, and tried to run the WRF 4D-Var.
Getting familiar with the new 4D-Var code. Will test the convective scheme using the new code when developing its adjoint.
The next meeting: 10 am 4/30/07, room 2072.