NCEP Synergy Meeting Highlights: June 3, 2013

This meeting was led by Wallace Hogsett (WPC) and attended by David Novak (WPC); Rebecca Cosgrove, Justin Cooke, and Chris Caruso Magee (NCO); John Derber, Vera Gerald, David DeWitt, Jun Du, Geoff DiMego, and Mary Hart (EMC); Joe Sienkiewicz (OPC); Kathy Gilbert (MDL); Andy Dean (SPC); Mike Brennan and Richard Pasch (NHC); Steve Lack, Steve Silberberg and David Bright (AWC); Bernard Meisner (SR); Jeff Waldstreicher (ER); Jeff Craven (CR); John Kelly and Aiyun Zang (NOS)

 

Special Presentation by Jun Du (EMC)

 

There are four SREF issues:

1.          Small vertical velocities in the NMMB members and zero vertical velocities in the NMM members. This fix is already in on WCOSS, and all members have vertical velocities that look realistic.

2.          The NMM members have a high dewpoint bias caused by GFS/GEFS IC (out of our control) and wet soil moisture. NDAS is drier than the GFS for the NMM. The dewpoints are now better at the initial time, but the bias comes back quickly (~9hrs). Replacing GFS soil moisture with NDAS will be tested in WCOSS parallel.

3.          The ARW members also have a high dewpoint bias. RAP ICs are now used on WCOSS. Because the RAP domain is smaller, there are some spin-up issues on the edges of the domain where GFS is used to fill in the gaps. The unbalanced IC issue is being mitigated by using a digital filter. The soil moisture is also too wet, and this is ongoing work. The ARW dewpoints are better with the RAP ICs, but the problem comes back quickly due to the underlying wet GFS soil moisture. Replacing GFS soil moisture with NDAS will be tested in WCOSS parallel.

4.          The effects of falling snow are causing the cloud ceiling heights to fall to the ground during snowfall. This is a post-processing fix that will be included in a quick-fix bundle.

 

The next SREF implementation isnít until Q4 2014, and #2 and #3 likely need a 30-day parallel, which isnít possible prior to WCOSS transition. Thus this will likely be addressed via a WCOSS parallel. It could be moved up before Q4, but itís a matter of prioritization. SPC: #1 is critical, but we understand that #2 and #3 rise above the bug-fix level.

 

1.      NOTES FROM NCO (Chris Caruso Magee)

 

The WCOSS experience can be painful, and NCO knows this. Tide has been down to fix bad cables, and Gyre is also going down to fix the same manufacturing issue. The service pack upgrade was not a magic bullet, and job failures are still occurring. IBM and the SPAs are tracking and addressing job failures. NCO is still trying to get a stable suite running and hasnít yet begun fine-tuning, with six weeks left before the deadline. NHC tested p-surge and hurricane models and had just a few glitches that were addressed. HYSPLIT is taking too long to run, but that may be fixed with fine-tuning.

 

NCO is working to streamline the supercomputer structure, with developers and SPAs working through SubVersion. This will help to install branches, but it means more up-front work between developers and SPAs. EMC wants faster implementations, and this streamlining could help. There has been discussion of getting a headstart on implementations by beginning the 30-day on the development machine and then continuing the 30-day period on the parallel supercomputer.

 

2.      NOTES FROM EMC

 

2a. Global Climate and Weather Modeling Branch (GCWMB) (John Derber)

 

Progress has been slowed by WCOSS instability. To clear up an earlier misunderstanding, the current GFS parallel is as close as possible to the CCS. Differences are coming due to round-off errors, bug fixes, and ICs, not data differences as previously reported. The random number generator used in the analysis is different, for example. Verification is close between the two systems, and it doesnít appear that any systematic differences exist.

 

NHC: WCOSS GFS does seem to spin things up less than the operational GFS, and that may be a good thing. It may also be less overly cyclogenic. Is it possible to see stats on the tropics? If not, we may abstain without a full evaluation.

 

EMC: Because of a TCVITALS issue, it hasnít been running stably for a long enough period of time to get stats. Thatíd scare me if I was evaluating.

 

CR Question: Does EMC ensemble the analyses? Is it better to do this?

 

EMC: Deterministic systems need a particular analysis. We donít run anything off the ensemble mean analysis. They need to evolve independently but we try to account for uncertainty through ensemble use.

 

2b. Mesoscale Modeling Branch (MMB) (Geoff DiMego)

 

The NAM bundle continues to evolve. Thereís a large reduction in the synoptic error. It has had a negative impact on temperature and 10-m wind bias in West and Alaska. There was an issue with WCOSS SSTs, which was fixed recently. Our stats for WCOSS NAM are better than CCS NAM so weíre comfortable with that solution. Every cycle we step back and pick up the global atmosphere and land state. Everything is running and weíre happy. Recently, we are testing a new initialization similar to the RAP that uses LHR to nudge toward during digital filter; that will be in the parallel. That will be part of the 2014 bundle. NAMX is performing better than the NAM.

 

2c. Marine Modeling and Analysis Branch (MMAB) (Vera Gerald)

 

Global RTOFS Atlantic: forecast extended from 6 to 8 days, late 2014 grib2 generation submitted, and an improvement to physics package will occur in early 2014.

 

For the Great Lakes wave model: new physics. Once WCOSS goes live, a parallel will be set up and a 30-day evaluation will be set up, and the upgraded model will be implemented in September or early 2014.

 

3.      NATIONAL OCEAN SERVICE (NOS, Aijun Zhang and John Kelley):

 

We ran successful tests on the Gulf of Mexico (GoM) nest on WCOSS and they looked good. NCO implemented more USGS stations that we need for GoM and Cook Inlet in Alaska. Thatís in /dcomdev and may go operational in early 2014. There will eventually be a SE and SW GoM domain.

 

We are running on WCOSS without any problems. New SF Bay forecast system is running on both WCOSS and CCS. We are working to help support Americaís Cup. The FY14 Schedule was sent to Becky.

 

4.      FEEDBACK FROM MDL/OPERATIONAL CENTERS/REGIONS

 

4a. MDL (Kathy Gilbert):

 

MDL collected data during May 9-31 for more systematic MOS output from GFS. Weíll focus on NDFD grids for CONUS and outside. Iím wondering if there are days that we should leave out? EMC would be uncomfortable using some of the days, but the only recent change has been tcvitals. That may change may affect the tropics, but it should be minor.

NCO will look to see when that was fixed. Would have been at the very end of month.

 

4b. NCEP Centers and NWS Regions

 

Weather Prediction Center (WPC, Wallace Hogsett):

 

WCOSS evaluation submitted to EMC. Big differences in GFS/GFS, minor otherwise.  

 

Storm Prediction Center (SPC, Andy Dean):

 

Thanks to NCO for all of their hard work.

 

National Hurricane Center (NHC, Richard Pasch): Nothing to report.

 

Space Weather Prediction Center (SWPC, no representative):

 

Ocean Prediction Center (OPC, Joe Sienkiewicz): Nothing to report

 

Aviation Weather Center (AWC, Steve Silberberg): Nothing to report.

 

Pacific Region (PR, no representative):

 

Alaska Region (AR, no representative):

 

Eastern Region (ER, Jeff Waldstreicher): Nothing to report.

 

Western Region (WR, no representative):

 

Southern Region (SR, Bernard Meisner): Nothing to report.

 

Central Region (CR, Jeff Craven):

 

It would be helpful to get hourly timesteps for first 12-24hrs from the NAM out to the field, as well as the GFS and others to support short-term forecasting efforts.

 

EMC: we do output from the NAM hourly to 49hrs; access can be had from NOMADS. There is a list to get into AWIPS. This is becoming more important. Weíll try to get this onto the list. Itís a distribution problem, but we agree it would be great for you to receive it.

 

5. NESDIS (no representative):

 

6. The next Synergy Meeting will be held at 2:30 pm EDT on Monday, 15 July 2013 in NCWCP conference room 2890, with remote teleconferencing capability.