September 10, 2009 Meeting Summary
Vijay Tallapragada discussed the 2009091006 failure of the operational HWRF for Linda (15E). The failure was a result of an issue with the SI and occurred because the boundary of the outer domain touched the dateline. While this problem is very unlikely to happen again, there is a slight chance it could. As a quick fix, Vijay suggested that NCO alter the initial storm location slightly, while he edited the HWRF scripts to account for such an issue. As a result of the 06Z failure, the 12Z run of HWRF for Linda was a cold start. NCO and TPC were contacted about this, and NCO re-ran the 06Z and 12Z cycles of HWRF for Linda.
Vijay also discussed the HWRF-DTC code merge status. While pre-processing and post-processing in the model are completed, there are issues with initialization and NetCDF files. The issues, as well as a few others, are likely to be resolved within the week and Vijay projected that testing on the merged code could begin next week with the hope of completion by mid- to late-October.
The issue of testing the HWRF and GFDL with the new GFS/GSI was mentioned, and it was decided that once a list of cases from the 2008 hurricane season was received from TPC, this could commence. To save time and computing resources, the entire 2008 hurricane season would not be run, but a representative subset of storms would.
Sam Trahan mentioned that HWRF code doesn't scale linearly on the processors on Vapor. This was attributed to the use of HWRF version 2 code, and is not expected to be an issue once version 3 code is in use.
Hyun-Sook Kim gave an update on her current work. The HWRF-HYCOM coupled system is being run in parallel for the 2009 Atlantic storms. So far, runs for Bill (03L), Danny (05L), and Erika (06L) have been completed with work on Fred (07L) ongoing. The 2009 version of HWRF is being used for these tests, which show the HWRF-HYCOM outperforming HWRF and H209 for weaker storms. In the near future, Hyun-Sook planned to work on a link between the ocean and HWRF website to show results of these tests.
Mingjing Tong said that she is currently running assimilations of tc-vitals MSLP data. She mentioned that her work involves inflating background error variances. The results did not show a significant improvement from this with enhanced variances along the storm path. Also, the pressure guess-fields showed a stronger storm than observed while the maximum wind showed a weaker storm than observed. This is likely due to an issue with the model's wind-pressure relationship, as discussed in previous meetings and in presentations given by Qingfu Liu. Mingjing plans to look at another case for this work. Mingjing also mentioned that Doppler radar data had been received for Ana and Bill from AOML. However, AOML was still working on transferring data from the airplane to ground format. John Derber elaborated on this issue saying that work on the Doppler radar data had previously been stopped. Thus, it would be a while before new data could be passed on to Mingjing. Assimilation of Doppler radar data would likely be something to work more on next hurricane season.
Qingfu Liu mentioned that he is still working with cumulus convection using the current SAS scheme and manipulating the cloud top to see if he could get improvement in the storm tracks.
Jun Du mentioned that his work with ensemble testing had halted as he waited for the new version of the GFS. He also mentioned that more personnel to assist in ensemble testing would be helpful.
Finally, Janna O'Connor gave an update on her work organizing HWRF meetings and keeping HWRF graphics up-to-date.