SR SSD 2001-25

Technical Attachment


Scientific Services Division
National Weather Service Southern Region

Fall 2001 Upgrade of the NCEP Meso Eta Model. An upgrade to the operational Meso Eta model is scheduled for November 27, 2001. This change package will increase the horizontal and vertical resolution of the four-per-day Meso Eta model runs and their supporting data assimilation from the current 22 km / 50 levels to 12 km / 60 levels. The package will also include the addition of NOAA-16 radiances to the three dimensional variational analysis (3DVAR) code and may also include direct use of WSR-88D radial velocities in the 3DVAR analysis, a new grid scale precipitation scheme and extension of the 0600 UTC and 1800 UTC runs to 60 hours.

The Phase II upgrade to NCEP's IBM SP computer in October, 2000 provided increased computational power that was initially used to perform real-time testing and evaluation of an 84 hour extension to the 0000 UTC and 1200 UTC runs of the Meso Eta model (see and of the 10 km nested "Threats" runs (see

The additional power now allows an increase in resolution of the Meso Eta model runs and their supporting data assimilation. This upgrade will provide higher resolution guidance for threatening local weather phenomena and the potential for more focused support of critical events such as fire weather and FEMA activities. The higher resolution will increase the likelihood that small scale weather features will be depicted in the model initial conditions which will, in turn, lead to better forecast guidance.

Model forecasts of surface winds and temperatures will be improved, especially where driven by complex terrain or coastlines (e.g. valley drainage and sea-breezes). More accurate forecasts of the evolution and intensity of threatening local weather phenomena such as heavy rains and associated flash flooding, heavy winter snows, lake-effect snows, local freezing temperature events and the location of the rain-snow line and drylines are also expected.

Depiction of small-scale features in the model's initial conditions depends on having observations of those features. The WSR-88D network provides a frequent source of truly mesoscale observations in the form of radial velocity and reflectivity. The latter are already used implicitly in NCEP's assimilation of Stage II analyses of observed precipitation (gages plus the radar estimates based on reflectivity). The addition of radial velocity data from the WSR-88D radar network will require little change to the existing 3DVAR analysis code. These data were first used during 10 km nested runs for the Atlanta Olympics in 1996. Use of these data involves processing of each radar's data into "super-obs" representing, for each available tilt angle, an average over roughly 5 km sectors for an hour's time. The NCEP has access to all radars through the NWS feed of the lowest four tilts with 4-bit precision. Refining the quality control (such as contamination by migrating birds) is the remaining unresolved issue with use of these data in operations.

A new grid scale cloud and precipitation microphysics scheme is being tested. Designed and implemented by Brad Ferrier. it has substantially more sophistication than the current simple scheme but with no appreciable increase in computer resources. The scheme distinguishes between cloud water and rain, and among ice crystals, rain, snow, sleet and graupel. It also determines different fall velocities of the various species of hydrometeors. The current scheme causes precipitation to reach the ground instantaneously, which has become a bad assumption as the model time step has decreased with increasing spatial resolution.

Eric Rogers has been making 12 km parallel runs of the Meso Eta model since 26 August and the results can be compared to the operational 22 km Meso Eta, the AVN and the NGM at Starting the week of 24 September 2001, 12 km Meso Eta model runs with just the NOAA-16 radiances and increased resolution will be made on the operational partition of the IBM SP. The final implementation version (including WSR-88D winds and/or the new cloud & precipitation microphysics) will be frozen no later than October 30, 2001.

The Web site has (or will have) a section on case studies and several will be referenced in the accompanying Technical Procedure Bulletin. A rerun of the Millennium Blizzard (30 December 2000 East Coast storm) will be made. A case of excessive QPF in both operational 22 km and Threats runs on 3-4 September produced a long north-south swath in Texas and which even fed back into jagged pressure and vorticity patterns aloft (Figure 1). When this case was rerun with the new gridscale microphysics scheme the amounts were drastically reduced and the pressure and vorticity fields were cleared of contamination.

The 12-36 hr QPF valid 1200 UTC 21 September was a definite winner for the 12km Meso Eta parallel run, which forecast more than twice as much precipitation over Oklahoma then the 22 km operational run (Figures 2).

The Mesoscale Modeling Branch would like to eventually add the off-time extensions of the Meso Eta to the AWIPS Satellite Broadcast Network and to increase the resolution of at least some of the critical surface grids. For now, the high resolution and extension grids are available on the NCEP server.

Anticipated impact on forecasts. Intense weather systems will be forecast more often. Oceanic cyclones have been deeper in the 12 km runs than the 22 km. The rate at which forecast error grows seems to be slower with the 12 km versus 22 km. The new gridscale precipitation scheme will eliminate the splotchy appearance of QPF that had been seen occasionally with the 22 km and continuously with the Threats and early 12 km runs.

Field Evaluation will be ongoing at NCEPs Hydrometeorological prediction Center, Marine Prediction Center, Storm Prediction Center and various NWS field offices through use of the 12 km parallel ETAX runs produced in real-time on the production side of the IBM SP. Western Region will evaluate the cold-season retrospective period of runs. Evaluations of the ETAX parallel with the combined change package will be included in the final form of the Technical Procedures Bulletin. Look for the first draft of this to appear on-line before the early November CAFTI meeting.

Future changes to the Meso Eta. By Spring 2002, the Mesoscale Modeling Branch expects to upgrade the Threats runs from nests of the hydrostatic Meso Eta system at 10 km and 60 levels to nests of a new (Janjic) nonhydrostatic Meso system at 8 km and 70 levels. This new model will use a hybrid vertical coordinate (terrain-following near the surface with constant pressure levels at and above 400 mb) instead of the Eta step-mountain coordinate used in the Meso Eta. All physics and a large amount of the operational Meso Eta model will remain the same in the new nonhydrostatic Meso model. Interested readers can follow the progress at:

January 2002 Upgrade of the NCEP RUC Model. An upgrade to the operational Rapid Update Cycle (RUC) model is planned for the week of January 21, 2002. This change package will increase the horizontal and vertical resolution of the hourly RUC runs from the current 40 km / 50 levels to 20 km / 50 levels. The package will also include the replacement of the optimum interpolation analysis with a three dimensional variational analysis (3DVAR) and will include the addition of many new observations to the analysis. The package will also include a new version of the Grell convective parameterization scheme, a new cloud analysis technique and higher resolution land use/soil type surface parameterization.

The 3DVAR analysis is expected to produce smoother analysis increments, a much better computational structure for using data such as satellite radiances and wind speeds, a better framework for sophisticated constraints (e.g., balance, parameterizations, full model), and better height/wind balance in analysis increment. More accurate forecasts should be generated from the more accurate initial conditions.

The frequency of model output will be increased to hourly for at least the first six hours of each model run, with output to 12 hours produced every third model run.

Anticipated impact on forecasts. Quantitative Precipitation Forecasts (QPFs) are expected to be more accurate, especially from convective precipitation events (Figure 3). Forecasts of in-flight icing potential are also expected to be more accurate due to the improved cloud and microphysics scheme. The new land-surface parameterization is expected to improve forecasts of precipitation and cloudiness and also produce a more accurate diurnal surface temperature cycle.

An on-line presentation of the new RUC-20 is available at: The URL of the RUC-20 discussion forum is: Model output graphics available at:

Other NCEP Model Changes. A list of the major model implementations planned for the remainder of 2001 and a preliminary schedule for 2002 can be found at:

As additional information about each implementation becomes available, links will be added on the change summary page.