MINUTES - NCEP Synergy meeting, February 26, 2001
Attending: Bill Bua, EMC/GMB; James Partain, MPC; Ed Danaher, HPC; Keith Brill, HPC; Geoff DiMego, EMC/MMB; Zoltan Toth, EMC; Steve Tracton, EMC; Pete Caplan, EMC/GMB; Hua-Lu Pan, EMC/GMB; Jim Hoke, HPC/MPC; Ralph Petersen, EMC
I. Status of IBM SP
Geoff DiMego unofficially reported for NCO. Benchmark tests on the "B" (development) part of the Phase II machine have been performed, with the resulting times a bit faster than initially expected, because of upgrades to the "load leveler" allowing for more efficient use of machine nodes. The "B" portion of the computer will be tested for 30 days starting late this week or early next week. While it is now available for use, the "B" machine may be down periodically until final testing begins.
Having both "A" (operational) and "B" parts of the Phase II available will mean doubling capacity, as to date we have been running both operational and development programs from an equally partitioned "A" part. However, this does not substitute for off-site backup operating capability.
II. Status of models from EMC
a. Global (H. Pan):
i. Current parallel run: Dr. Uccellini has been briefed on the new AVN/MRF model physics, including predicted cloud water and cumulus momentum mixing, and he has given the okay for parallel testing and NCEP operational center assessment to begin. The Global Modeling Branch (GMB) is looking at a March review by the operational centers (including retrospective runs), followed by presentation for approval by CAFTI in April, for operational implementation any time from late April to June.
ii. Planned new parallel run: The GMB is now planning for its T264L64 runs after the implementation of the current parallel with its new physics (beginning in May/June). These will be available for in-depth examination by the Tropical Prediction and Hydrologic Prediction Centers. There are no plans currently to increase the resolution of the AVN/MRF output grids from their current 1° resolution to 0.5° with this resolution change.
b. Mesoscale (G. DiMego):
i. 84-hour extension: The 84 hour extension will be run in real-time for test and evaluation. All products currently available except for BUFR soundings will also be available in the 84-hour extension. The 35 minutes between the ending of the 60-hour Eta and beginning of the AVN run will (barring delays) be more than adequate to fit in the 60-84 hour forecast. If there is a delay, a slower executing (on fewer nodes) program will be used to run the Eta extension concurrent with the AVN, with no effect on the AVN delivery time and some delay in completion of the 60-84 hour Eta forecast.
ii. New SST problem: A problem in the new OMB SST analysis was noticed earlier this month, where an anomalously warm pool of water appeared off the Mid-Atlantic coast. Given the coastal cyclogenesis problems noted earlier in the winter resulting from SST analysis problems, an emergency change was put in to remove a climatological increment used when real-time observations are not available to update the SSTs. The Eta was recycled from the adjusted SSTs, with no negative impacts noted.
iii. 3D-VAR data assimilation problem: Dave Parrish is almost ready with a test version of a modified 3D-VAR that uses a initial balance constraint more similar to the AVN/MRF's SSI balance constraint (on divergence tendency calculated by model dynamics) rather than the too strict thermal wind constraint currently in use. This will be put in the new parallel for spring implementation (see below)
iv. New Eta parallel: The new parallel will include the 3D-VAR change above, a retuning for the change in seasons, and a change in the code that will make the data assimilation system more readily tunable. Other changes include assimilation of observed precipitation into the 3D-VAR analysis and an upgraded land surface parameterization. The parallel run will be put in near-real time, with warm and (hopefully) cold season retrospectives for NCEP center-wide review before a hoped-for April or May implementation.
v. 12-km Eta: After the "B" machine 30-day acceptance test, test runs of the higher resolution Eta will be performed using idle production nodes. Grids will be made available via ftp to view test run fields, and an implementation team will be put together to facilitate review of the 12-Km Eta. In response to a question on the effect of the SSTs at the higher resolution, Geoff DiMego responded that he thought the 12-Km would behave similarly to the 10-Km "Threats" nests, except for any changes that might result from the 12-Km Eta cycling on itself, rather than getting initial conditions from a coarser model.
vi. Grid/Visualization on NAWIPS/AWIPS: There was discussion of limitations of availability of data and its visualization on AWIPS. An 88-Km coarse resolution AWIPS grid will replace the NGM 90-Km grid to get the "big picture" from the Eta; it is thought that the finer details would only be necessary on more local domains that could be picked up from different sources (e.g. the Eta tiles).
a. Short-Range Ensemble Forecasts (SREF): Steve Tracton reported that real-time testing of the SREF system would begin upon approval by Dr. Uccellini subsequent to a briefing on Friday, March 2, 2001. These will be run at 48-Km or equivalent resolution from off-time 09Z and 21Z initial conditions. GRIB and metafiles will be created on the IBM SP including ensemble mean and spread, and probabilities of exceeding precipitation thresholds (same information as is currently available on the SREF web page). Data from individual ensemble members will be available as well. These data will be made available on NAWIPS.
A final product list is under development and will be presented as soon as available.
b. Global Ensemble System:
i. Files will be made available in GRIB for currently available data found on the web. This should become available in several weeks.
ii. Post-processing changes to be implemented in the next 3 months or more include removing model bias from the mean and spread.
iii. Changes in determining the initial perturbations are in the process of being made to make the timing of the global ensemble runs match that of the relocation of tropical vortices spun up in the AVN/MRF. In particular, we are considering implementation of a 6-hr (instead of 24-hr) breeding cycle.
iv. Increased computational resources will be put to use by testing the value added to the ensemble forecast from adding ensemble members and increasing ensemble resolution from T62 to T126 in the 3.5 to 7 day range. Value will be considered added if the ensemble forecasts produce a smaller spread, with an increased likelihood of the verifying analysis appearing within that spread.
v. Question arose regarding usefulness of 12Z ensembles. For some applications (e.g. Ocean Prediction Branch), the 12Z ensembles are useful for the middle range forecasts, which they produce at night. For the HPC, however, the ensemble forecasts are not currently available soon enough for careful evaluation before its forecasts must be issued.
IV. Next meeting time and date: The next meeting is tentatively scheduled for Monday, March 26, 2001 at 12 noon.