4DVAR minutes 013107
From MmmWiki
10:00a, room 2072
Participants: John M., Tom, Hans, Yongrun, Michael, Jimy, Qingnong, Xiaoyan, Jenny
Contents |
Project Scientist I
Hans informed the group a new PS-I was just advertised. He/she will work 30% on EUMESAT (this year, and later 4DVAR), 20% on 4DVAR, and 50% on hybrid.
AFWA and NPOESS 2006 final report
Hans made a proposal for the writing.
AFWA and NPOESS 2006 WRF 4D-Var final report AFWA (1.45) and NPOESS (1.00FTE) First draft: Feb 15 2007. Hans, Jimy, John M, Qingnong, Tom, Xiaoyan, Xin Revised version: Feb 28 2007. Hans a) Development of simplified physics (AFWA 0.25, NPOESS 0.25). Qingnong, Jimy, Xiaoyan and Xin b) Coupler and Parallelization (AFWA 0.75, NPOESS 0.3). John M, Tom, Qingnong, Xin c) Weak constraint for noise control (NPOESS 0.45). Xin, Hans d) Meteorological tests (AFWA 0.45). Hans, Xin
Qingnong, John M and Hans agreed to take the lead of writing a), b) and d). Xin agreed earlier with Hans to lead the JcDFI writing.
However, John asked Hans to check whether it is necessary to write such a report. We need to prepare for the final review. A PPT may be preferred. Hans agreed to check the SoW and to discuss this further with Dale.
Progress review
Physics
Xin has merged the PBL and LSC into WRFPLUS, which is a part of the AFWA release package.
PBL: Veritical diffusion
NL, TL and AD have all passed test.
The scheme has a significant impact on nonlinear runs, but only marginal impact on 4D-Var. Furhter 4D-Var tests with higher resolutions may be needed.
LSC: Large Scale Condensation
TL and AD have been developed, but the test still has problems.
There are two additional, unexpected subroutines which also need TL and AD. This delayed the work somewhat.
TAF
The source-to-source converter works fine now and produces code which gives the bit-by-bit correct results compared to the original nonlinear code.
However, TAF still fails to generate correct TL and AD on the solver level. Before Chrismas, Tom made a testing setup for Ralf and Thomas of FastOpt on their request. Recently, Tom made a "one-button" setup for them. Although Tom got a few newer versions of TAF, none produces correct results.
The recommendation: we cannot rely on TAF for automatically and regularly generating TL and AD from new versions of NL WRF. It is still useful on subroutine level and can save manpower in the convection task and J_lbc task, planned for 2007.
Tom and Qingnong agreed to run OPENAD on LSC to compare the results.
Meteorological testing
In the past month, major effort was on the preparation of the AFWA release code. Now the 4D-Var related merging is finished. The LSC code is still under testing.
The single processor WRF 4D-Var for release is almost done. The benchmark experiments are under review. Once done, we will move to multi-processors experiments.
AFWA 2007 work plan
Hans presented the first draft of the proposed AFWA 2007 4D-Var work plan.
| Priority | high | medium | high | high | low | low | |
| Multi-inc | J_lbc | Optimize | Met. test | convection | A2C | Sum | |
| Dale | 0.15 | 0.15 | |||||
| Hans | 0.1 | 0.1 | 0.1 | 0.5 | 0.80 | ||
| PS-I | 0.2 | 0.20 | |||||
| Qingnong | 0.3 | 0.30 | |||||
| SE | 0.3 | 0.30 | |||||
| Xin | 0.25 | 0.25 | 0.25 | 0.75 | |||
| Sum | 0.35 | 0.35 | 0.65 | 0.7 | 0.3 | 0.15 | 2.50 |
Hans explained the priority he set and some arguments behind the numbers.
A further discussion among Dale, John M, Tom, Xin and Hans is needed for the optimization task.
John M suggeted to include a task for day-to-day code coordination and testing.
Qingnong considered 0.3 for developing NL, TL and AD of a simple convection scheme to be too low. 0.4 at least. 0.5 would be preferred.
The work plan will be iterated a few times, with Dale's involvement, in the coming month.
AOB
The next meeting: 10:00am, 2/28/2007.
