SR/SSD 99-14

Technical Attachment

Highlights of CAFTI Meeting - April 15, 1999

Scientific Services Division
NWS Southern Region Headquarters

NGM and NGM-MOS Beyond Jan. 1, 2000: Reversing earlier plans to terminate the NGM model this year, NCEP will convert the NGM code to run in one-way nested mode on a single processor of the IBM SP (the new Class VIII computer). Initial conditions will come from the Eta analysis, with boundary conditions from the six-hour old AVN (e.g., boundary conditions from 0600 UTC AVN run for 1200 UTC NGM run). This will allow the continuation of NGM-based MOS, and TDL has agreed to convert the MOS equation code to run on the IBM SP. Parallel runs (current NGM vs IBM SP version) should start within a month to confirm no significant degradation in MOS output. Plans now are to continue the NGM at least through April 2001, but NGM MOS will eventually be supplanted by MOS from the AVN model (and/or possibly the Eta).

Development of AVN-Based MOS: TDL continues development of the AVN-MOS suite. The use of the AVN model for MOS ensures output for all NWS sites (contiguous 48 states, Alaska, Hawaii, Puerto Rico, and even Guam!). A critical issue, particularly with the Eastern and Southern Regions, is the arrival time of MOS output for offices in the Eastern Time Zone. The regions indicated MOS output must be available at field offices by 1630 UTC for use in morning forecast package. With the current NCEP model job streams, AVN MOS cannot meet that deadline, which is why emphasis has been placed on continuation of NGM MOS. NCEP and TDL, working with the regions, will develop a list of possible solutions. For example, to compensate for the eventual loss of NGM MOS, the AVN run could be started earlier, or MOS could be provided from the 0600/1800 AVN runs rather than (or in addition to) 0000/1200 AVN MOS. CAFTI to consider will consider the options.

Approved Change in Eta Analysis Scheme: Parallel runs of Eta (at 80 km resolution) indicate that improved tuning of 3DVAR parameters (essentially giving more weight to wind reports and less weight to temperature reports) produces better analyses and forecasts of heights, winds and temperatures, and perhaps some improvement in precipitation amounts. CAFTI gave provisional approval for this change, pending the field offices reviewing the statistics and the draft TPB. Documentation of expected impacts to model output will be added to the NCEP model Web site. Current information is available at:

Anticipated Change in AVN/MRF Resolution: Parallel runs of the Global Spectral Model (used for AVN, MRF and Global Data Assimilation System) for a five-week period last summer and a three-week period this winter show the problems encountered when the model resolution was changed from T126L28 to T170L42 last summer have been fixed. NCEP proposes restoring the operational model to the higher resolution at the time it is ported to the IBM SP later this year. The corrected T170L42 version demonstrated better skill than the T126L28, especially over North America. Skill over the tropics was somewhat worse. The upgrade of the resolution should produce little change in large-scale circulation patterns, but NCEP expects to see deeper storms in the tropics as well as in the mid-latitudes. Precipitation forecasts should be slightly more accurate. Other expected impacts include:

Better fit of the model initial conditions to the observations.
Better short range forecasts.
More intense mid-latitude and tropical storms.
Better resolution of jet streams.

Development of spurious tropical storms in the 36-72 hr period remains a problem. However, NHC believes the improvements in the 0-24 hr period warrant the change. A draft Technical Procedures Bulletin describing the proposed change is available at:

(Links to model descriptions, model changes and model output can be found through the SRH Numerical Weather Predictions Links Web page: