The NGM is a 16-layer model with 80 km resolution. It generates 48-hour forecasts twice a day and is used for model output statistics guidance. Development has been frozen on the NGM since 1990 and the model itself will be discontinued in 1998. This is after the Class VIII system and the model enhancements that it will support are in operational use. The NGM forecast hours are 00-48, running at 00Z and 12Z.

The early ETA run involves a 38-level model with 80 km resolution. Like the NGM it generated 48-hour forecasts twice per day. The ETA utilizes a first guess from the global data assimilation system (GDAS) and an ETA optimal interpolation (OI). The need for the early ETA run will be obviated by the aviation (AVN) global model when it begins to fun four times per day in 1996. Once the AVN demonstrates accuracy comparable to or better that the ETA for precipitation guidance in the 24-48 hour range, then the early ETA will be discontinued.

This OI-based data assimilation system is used to initialize the mesoscale ETA model. It is performed on the model domain at 29 km resolution and for 50 levels. It will become operational in early 1995 and will initially run two 12-hour cycles per day, each with an intermittent (3-hour) assimilation of data. Later in 1995 the EDAS will begin to run four times per day. In 1998 the EDAS will evolve into a four-dimensional variational regional data assimilation scheme based on the adjoint of the ETA model. It will then run four times per day with continuous data assimilation over 6 hours for each of the four cycles.

This is a 50-layer model with 29 km resolution. Initial conditions come from the EDAS. The Meso-ETA will become operational in early 1995 and will produce 33-hour forecasts twice per day. The base times for these runs will be 03Z and 15Z in order to use updated boundary conditions from the AVN model. Later in 1995 the Meso-ETA will begin to run four times per day. Special ETA forecasts will be run over Alaska if computationally feasible. After the arrival of the Class VIII system, plans are to consider an increase in the horizontal resolution of the Meso-ETA to 15 km and to increase the number of levels to 70.

This model is presently under development. The RSM, now run experimentally within the global spectral model , is a 28-level model with 40 km resolution covering a North American sub-domain of the NGM. It produces one 48-hour forecast per day for comparison with the ETA model. The RSM should become operational in 1995, perhaps with an increase to 42 levels. After the operational implementation of the Class VIII system in 1998 the RSM will run with 20 km resolution within the AVN model four times per day as an enhancement over North America. In the same time frame a non-hydrostatic version with 10 km resolution will come into use experimentally.

The Quasi-Lagrangian Model (QLM) is currently operational. This is an 18-layer model with 40 km resolution. It produces 3-day forecasts when needed. The GFDL Multiply-nested Movable Mesh (GMMM) model, also an 18-level model but with 20 km resolution on an inner grid, is now running experimentally. The QLM will probably be discontinued in favor of the GMMM in 1995. The GMMM will increase its resolution to 10 km when it moves to the CLASS VIII system in 1998.

The RUC is a hybrid sigma-isentropic analysis and forecast system. It utilizes a 3-hour, OI-based data assimilation cycle to produce 12-hour forecasts eight times per day. This is a 25-layer model with 60 km resolution. With the implementation of the Class VIII system in 1998 the will run hourly, 24 times per day. The resolution of the model will increase to 50 levels and 30 km in that same time frame.

The outlook for the 5 years after 1998 is that, the NCEP will run a national domain mesoscale model at 5-10 km resolution based on a non-hydrostatic version of the ETA model (possibly the RSM), together with ensembles of lower resolution. This system will be coupled with multiple storm-scale (1 km or less) models used for very short range forecasts (up to 6 hours). The storm-scale forecasts will be nested within the national domain in areas threatened by severe weather. Together with extensions of NCEP models, candidates for such storm-scale models could be those developed at universities or other research centers. Candidate models will be evaluated for potential operational use in the Model Test Facility to be established within the Environmental Modeling Center.

The AVN is a 28-level model with 100 km horizontal resolution which runs twice per day producing 72-hour forecasts. It will be refined to 40 levels and approximately 80 km resolution in late 1994. Another advancement will be the conversion from spectral to semi-lagrangian methodology in 1995. The AVN will run four times per day beginning in early 1996. In 1998 the AVN will run with 60 km or finer resolution using 50 levels. The RSM will run within this model at 20 km resolution providing the main synoptic guidance over North America.

The MRF is the same numerical model as the AVN but it will run under a wider variety of conditions and for different purposes. By late 1994 it will run once per day at the resolution of the AVN to produce a 7-day forecast. A low-resolution version of the MRF (using a 2.5 degree grid) will run out to 16 days as a tool for ensemble forecasting. The low-resolution MRF will also run experimentally on the Class VII machine with an imbedded version of the RESM as referred to above.

In 1998 the MRF will probably be used in ensembles of a few runs with 70-100 km resolution out to 7 days; and ensembles of many low-resolution runs for days 7 to 30.

The encouraging experience thus far at the NCEP with the ensemble system presently available for the 1-5 day global forecasts suggests that this system can be improved and extended to longer ranges. The use of ensembles of forecasts has already begun and will provide a quantitative foundation for probabilistic synoptic forecasting beyond the first week. Even for short-range forecasting the use of ensembles will provide guidance for the probability distribution of precipitation and extreme events.

The GDAS will continue to use the spectral statistical interpolation technique but will add refinements as they are developed. By 1998 it will apply four-dimensional variational methods with a 12-hour assimilation interval and continuous data utilization.

The GDAS will also use a number of new or expanded data sources. Satellite instrumentation in particular will provide new data that will be incorporated by the GDAS over the next several years. Data from profilers and from new aircraft sensors will be assimilated too. In addition, the GDAS will be refined to better utilize information from traditional sources such as pressure tendencies and cloud reports from surface observations.

The outlook for NCEP global models after 1998 is that they will have a horizontal resolution of about 20 to 40 km and will run through 72 hours in the case of the AVN and through 7 days for the MRF. Ensembles of somewhat lower resolution model forecasts will be run for one month or longer. With better data assimilation, skillful tropical forecasting based on these models and ensembles may become possible through 10 days. Ozone forecasts should be routinely available and forecasts of atmospheric contaminants may also prove useful.

The third generation WAM (WAve Model) global model will begin running on the Class VII system in late 1994 replacing the NOAA Ocean Wave (NOW) model. It will run twice a day out to 72 hours with 2.5 degree resolution. By 1998 NCEP will utilize a nested grid ocean wave model on the Class VIII system. The resolution of this model will be one degree in the open ocean and 0.25 to 0.5 degree in selected regions along the East and West Coasts, the Gulfs of Alaska and Mexico, and in the area around Hawaii.

In July 1994 a prototype Mellor model and data assimilation system will begin to run routinely for the East Coast of the United States. If this experiment is successful, it will become operational in 1996 and a similar experiment will begin for the West Coast.

In the out years of 1999 and beyond, high resolution (5 km) coastal nowcasting and short-range forecasting should provide guidance to fisheries, ships, and environmental monitors along the entire North American coast with forecasts of thermal structure, currents, and water levels, including tides and wind induced storm surges.

The wave models over the coastal domains will be coupled to the CODAPS models to take into account wave-current interactions which significantly modify the waves.

The reanalysis system will use 28 levels and 2.5 degree resolution. It will reanalyze 35 years of data, performing at a rate of 30 days of analyses per calendar day. This will proceed operationally beginning in late 1994. After the implementation of the Class VIII system in 1998, another reanalysis cycle will begin with a domain of approximately 40 levels and one degree resolution.

This program is designed to monitor the climate and climate change. The CDAS will become operational coincident with the reanalysis system. It will assimilate current data at the same resolution as the reanslysis program. The CDAS and the reanalysis system will both run on the same computer system.

In the period after 1998 a long reanalysis with a state-of-the-art (but frozen) date assimilation system will be performed about every 5 years. This will provide guidance to climate monitoring and forecasting efforts. It will also be valuable to compute "perfect prog" statistics, linking model analysis with station observations using an extremely long training period. This can be combined with adaptive model output statistics in which the model forecast is used to predict the analysis, requiring short training periods.

The COAM uses an 18-level version of the MRF model with approximately five degree resolution. This is coupled with a Pacific Ocean model to produce long-range forecasts. The Coupled Model Project will use the Class VII system to make on 12-month coupled model forecast per week. This forecast permits dynamic interaction between the atmosphere and the ocean, a process that currently results in significant systematic biases in the forecast of sea surface temperatures. These biases are corrected in a post-processing step. The Coupled Model Project will also produce two 6-month forecasts each week from the atmospheric model alone using the corrected sea surface temperatures from the coupled model run. These forecasts are both for the same 6-month period and are run from different initial states.

By late 1995 the COAM will be extended to a 28-level 2.5 degree resolution MRF/Pacific Ocean model.

*Information provided by NOAA/NMC-1/95*

**(all references to dates are approximate and reflect current estimates
as to the availability of advanced computing systems and the results of diverse
research and development activities)**