Summary on CWB project meeting on 27 April 2005

Notes on CWB project meeting on 27 April 2005

Attendants: B. Kuo, Y.-R. Guo, H.-C. Lin, X. Ma, S.-Y. Chen

Discussion on Task 1:

Retrospective experiment

Hui-Chuan: CWB selected the month of July 2004 (summer) and January 2005 (winter) as the experimental months. Exp for July 2004 with the current WRF 3D-Var has been carried out. The preliminary comparison between 3D-Var and operational indicated that the 3D-Var Exp, in general, is worse than operational. However, this comparison may not be fair at this moment.

a) verified against the different analysis: 3D-Var/ forecast to 3D-Var analysis, and NFS/forecast to NFS analysis;

b) Verification made on pressure levels, so there are many vertical interpolations for 3D-Var/forecast.

However, both 3D-Var and operational verified against CWB global analysis have been done. 3D-Var Exp showed serious diurnal oscillation in errors.

Bill Kuo : Can we use the month of 15 Aug. to 15 Sept. 2003 for Retros. Exp? Then, we have the same month for Typhoon Dujuan Exp.

Also can we consider to use EC analysis in verification?

Y.-R. Guo : Verification against the observations, such as soundings, is also necessary.

TC Bogus

Y.-R. Guo: CWB TC bogus data generator, decoder, and 3D_Var OBS preprocessor program have been developed and tested for 2003062912Z data. Need the TC Bogus data for Dujuan (2003083012Z) for 3D-Var run test.

Bill Kuo : What approach to generate TC Bogus? Need document.

Global bogus

H.-C. Lin : Global bogus data should be ready in CWB, just need to test with the 3D-Var system.

(After meeting, I (Y.-R. Guo) wonder where the global bogus are deployed, just over ocean? According to what regulation in generating the global bogus in CWB?)

New NFS BES

Y.-R. Guo : New gen_be code and testing data have been sent to CWB a month ago. But the conversion between (u,v) and (y,c) still has problem, need to figure out and fix it.

H.-C. Lin : No problem to run gen_be code with the testing data in CWB. Need to separate the Stage0 code from the 3D-Var code, then we can start CWB NFS data work for gen_be.

Y.-R. Guo : Does CWB consider to develop a “directly transfer NFStoXb”, like KMA’s KMAtoXb? This approach will avoid the interpolation from NFS s to WRF h in vertical and some unnecessary variable’s transformation. (Bill: Could you talk Dr. Terng for this when you are in Taiwan? The interpolation to and back, transform to and back between NFS and WRF may introduce the errors)

H.-C. Lin : no this plan yet.

Y.-R. Guo : If not, CWB could prepare the NFS forecast data files for a month (15 Aug. to 15 Sept. 2003) in WRF netCDF format by using existed H.-C. Lin’s converter program with 12-h and 24-h forecast in each of files. Then, the current gen_be can be directly applied to CWB NFS data.

H.-C. Lin : Hope that Stage0 code separate from 3D-Var code, which will make things easy.

Bill Kuo : Dale Barker will give the training lectures in early June when he visits Taiwan. Why Y.-R. let X. Ma run WRF forecast for a month?

Y.-R. Guo : Let X. Ma to be familiar with WRF model system and NCAR computing facilities. Also the BES from WRF forecast could be served as a benchmark to check CWB NFS BES.

Discussion Task 2:

GPS RO observation error

Y.-R. Guo : Shu-Ya has done much work to look at GPS RO data from COSMIC/CDAAC. The decoder programs converting the data to LITTLE_R format are already there. But X. Ma should use wetPrf data with a lower vertical resolution for the local operator assimilation. X. Ma should be familiar with CDAAC data and make a little bit modification for the decoder program.

The observation error specification in the 3D-Var OBS preprocessor can be found from Kuo et al (2004). The errors are time-, geographic-, and altitude-dependent.

Bill Kuo : I have reprint of that paper for X. Ma and H.-C. Lin. Why we cannot use GPS RO BUFR data? AFWA, etc. like BUFR data?

Y.-R. Guo : Currently we do no have person familiar with BUFR format. Yes, Dale talked me about work on BUFR data in a future plan.

GPS RO quality control

Y-R. Guo : Currently, just a 5s check with the innovation in 3D-Var. First, we may implement the Lidia Cucurull (JCSDA) procedure in WRF 3D-Var, then tune it through experimental runs.

Bill Kuo : We may also incorporate something from Martin Lohmann’s work.

Improvement of the GPS RO local operator

Y.-R. Guo : To consider the vertical variation of the perigee points, the preprocessor has also included the position of the perigee points in each level data record. However, the 3D-Var code needs to do corresponding modifications.

An accurate refractivity index formulation will be used to replace the current one. When I worked on the ZTD assimilation, I found that the current one has some unnecessary approximations.

Typhoon Dujuan experiments

Bill Kuo : Having all improvements implemented, when will we conduct the experiments for Dujuan?

Y.-R. Guo : August is OK?

Bill Kuo : We should have meeting once a month. We may have the next meeting late May. In August, there is a conference on application of GPS data in D.C.. Do you have been D.C. before, Hui-Chuan?