4DVAR minutes 09/29/06
10:30 am 9/29/2006
Lilly room
Participants: Chirs, Dale, Hans, John B, John M, Qingnong, Tom, Xiaoyan, Xin
Contents |
Progress review
Merge 4D-Var branch to the main trunk
John B stopped the work on the wrfplus of the main trunk. Xin will continue the work. The deadline is moved to Nov 30.
Veritical diffusion
Qingnong and Jimy discussed the issue. Jimy is working on a simplified vertical diffusion scheme. Testing results will be reported next time.
Nup and nup_inv
No progress.
Coupler and parallelization
The parallel code can run. John M worked further on optimization of the code. However the adjoint model does not scale well and John M is working on the issue. An efficient parallel 4D-Var should be ready at the end of the year, Dec 31. Xin has been and will continue to be involved in the testing. Xin also merged John’s new code to the 4D-Var branch and will this merging work (now to the branch and later to the trunk).
Tom will need another two weeks to reach a recommendation for TAF. He is close to make a full set of tl and ad using TAF. Dale and Hans will talk to Ralf in the coming adjoint workshop to find out more about TAF. Tom also had a look at the French tool.
Meteorological testing
The 4D-Var scripts have been further developed to make the “centered” assimilation window. They are re-running all the cases around typhoon Haitang plus a cycling run.
Large-scale condensation
No progress. The deadline Dec 31 should be kept.
JcDFI
JcDFI has been implemented with QN and some preliminary results indicated that it works as expected. However its impact on meteorological parameters will be investigated.
JcDFI does not work with CG. Xin will discuss the issue with Rizvi.
4D-Var versions
Hans proposed a set of version numbers for 4D-Var. John B suggested to name the version we plan to deliver to AFWA and KMA 2.2 to be consistent with model and 3D-Var.
Version 1.0. Jan 2006. The prototype. Delivered to AFWA Jan 2006.
Version 1.1. Jun 2006. CV5, more obs, scripts, testing datasets.
Version 2.0. Aug 2006. First parallel version.
Version 2.1. Oct 2006. JcDFI implemented. Optimization. New scripts.
Version 2.2. Dec 2006. Optimized and parallelized version using 3 executables. Vertical diffusion and large-scale condensation included. Scripts and testing dataset. Simple documents.
AOB
The next meeting: 10:00am, 10/31/2006.