EMC Synergy Meeting Highlights - 04/24/00
Rapid improvements and modification to operational models and the platforms the models are run upon requires maximum synergy and frequent interaction between the modeling community and the operational community within the national weather service (NWS). This is needed in order to achieve optimal model performance and accelerate model development consistent with operational need.
Therefore, in order to meet these goals, monthly meetings are attended by the environmental modeling center (EMC) and the co-located operational centers including the Hydrometeorological prediction center (HPC), the Marine Prediction Center (MPC), and NCEP Central Operations (NCO) of the World Weather Building (WWB) in Camp Springs, Maryland.
THE FOLLOWING ARE HIGHLIGHTS FROM THE MOST RECENT MEETING. INFORMATION FORWARDED IN THESE HIGHLIGHTS ARE TO BE REGARDED AS TENTATIVE AND FOR PLANNING PURPOSES ONLY. THE INFORMATION FORWARDED IN THESE HIGHLIGHTS ARE SUBJECT TO CHANGE AND SHOULD NOT BE RELIED UPON OR REFERENCED AS FIRM DEADLINES OR ESTABLISHED POLICY.
This meeting was led by Peter Manousos (SOO - HPC) and followed a generic agenda denoted by the subtitles below. Attendees included Carl Staton (NCO), Dave Reynolds (Forecast Operations Branch Chief - HPC), John Ward (NCO), Dr. James Hoke (Director - HPC & MPC), Kevin McCarthy (Deputy Directory HPC), Dr. Stephen Jascourt (COMET Post Doc. at HPC), Hua-Lu Pan (EMC - Global Modeling Group), Geoff DiMego (EMC - Mesoscale Modeling Group), James Partain (SOO - MPC), Bill Bua, Edward Danaher (Development Training Branch Chief - HPC), Keith Brill (Research Meteorologist - HPC), Brett McDonald (Research Meteorologist - HPC).
III. Meeting Highlights
1. IBM SP
a. Current Status: John Ward stated that the IBM SP has been fully operational since 3/27/00 and there are no dependencies on the J-90 machines. Although not mandatory to utilize, two to three hours per month of down time has still been allocated for maintenance in the case where any single points of failure need to be serviced on the SP. There is also a plan to take the ENTIRE Bowie complex off line as GSA needs to perform "switch gear" work. Tentatively this would occur over a 3 day weekend in the late fall of this year after the upgrade of the IBM SP to PHASE II has been completed. For reference, PHASE II on the IBM SP is slated to be delivered in September of 2000 and includes an approximate doubling of disk space, an increase from 2 processors per node on 384 nodes to 16 processors per node on 128 nodes, and an increase of processor speed from 300MHZ to 375MHZ.
b. Planning of PHASE II IBM SP: Carl Staton stated that 3000 square feet at the Bowie complex had been allocated for NCO use in preparation for upgrade of the SP to PHASE II and that this was originally intended to be available to NCO in August 2000. However, the Census Bureau is currently utilizing this space and will not vacate until November 15 2000. Therefore, options are being explored by both IBM and NCO to facilitate the upgrade of the SP to PHASE II.
c. Potential Impact of scheduled 4/26/00 DBNET upgrade on IBM SP on WWB Operational Centers: None foreseen since WWB operational centers use DBNET outside of the IBM SP loop.
2. Status of Ongoing parallel/testing efforts from EMC
a. Global Modeling Group: Hua-Lu Pan stated that outside of the changes to the convective scheme there has not been much difference observed between the MRFx and the operational MRF model. Some minor differences observed include a better equitable threat score for QPF at the .01" and .10" thresholds and some improvement in the tropics in the anomaly correlation scores of winds. The implementation of the AVN to run to 126 hours is planned for May 16 2000 and the APO has assured NCO that AWIPS 4.3.1 will be able to properly ingest and display the grids from this model. This sets the stage for the GFDL hurricane model to be run to 120 hours by June 1 2000. Further, it was also stated that the NOAA WaveWatch III model will also be run out to 120 hours within a few weeks of the May 16th implementation of the 126hr AVN.
b. Mesoscale Modeling Group: Geoff Dimego discussed the 2 current parallel runs at 48km resolution. The first was designed to test the neural network of SSMI data to obtain more winds where there are clouds. The result of 2 weeks of testing yielded worse verification scores than the control. The second tests the direct use of radiances utilizing new 3DVar code. This is still being evaluated but is expected to be a success and implemented in August when the Eta horizontal resolution will be increased to 22km. The next planned parallel is in construct and is 3 weeks away from being run in parallel mode. This will also be run at 48km resolution and will use the observed precipitation analysis as part of the assimilation system. This will also potentially be included with the August bundle.
3. Input to EMC from Operational Centers
a. Dave Reynolds queried Geoff Dimego as why the Eta model is showing poorer verification scores than the AVN and NGM (low bias) at the 1" or greater QPF thresholds (which is similar to last year's verification scores). Geoff theorized that it likely has to do with the manner in which horizontal diffusion and damping of divergence is applied. These are in effect "noise control" measures but may be too "heavy handed" because they were originally tuned for longer time steps than are currently used. Since the time steps have been shortened to accommodate increased horizontal and vertical resolution and the diffusion and divergence damping were NOT formulated to be time resolution dependent, there is too much "noise" removal with the current tuning parameters. It is realized by the Mesoscale Modeling group that if this goes unchecked, it will become a more predominant source of error with the increase of the Eta horizontal resolution to 22km.
b. Mike Davison's comments (from the International Desk at HPC) were forwarded on his behalf which indicated a strong dislike for the performance of the MRFx. Particularly, Mike described a tendency of the MRFx to over forecast precipitation amounts in both the tropics and with frontal systems, as well as a tendency of the model to shift the position of the Bolivian High toward regions where the model has erroneously produced enhanced divergent areas in association with over forecast precipitation in the tropics.
Peter Manousos forwarded a question from the field to the meeting attendees which asked if it was possible to centralize or consolidate all the NWP-related information available from HPC, EMC, etc. to one location. The group's consensus was that this is possible, however this would be an enormous endeavor for any one individual to undertake. The most feasible solution is to utilize one of the "summer-hire" employees HPC obtains for this task.
IV. Next Meeting Proposed June 5, 2000 at noon in room 209.