INTERACTIVE FORECAST PREPARATION POSITION PAPER 3


Overview of "Model-based Initialization":

Generating Initial Surface Forecast Values from NCEP Models

Stuart Wier


NOAA Forecast Systems Laboratory


Boulder, Colorado


July 3, 1997; August 22, 1997


SUMMARY

Numerical model output can be used to make initial values of all weather elements used by the IFP system. Values are made for all grid points in the grid used by each office's IFP. The model-based grids can be made anywhere inside each model's native grid, which is usually the entire AWIPS 211 or 212 grid, covering all the lower 48 U.S. states and adjacent waters and lands. Adjustments are made for terrain elevation so that derived elements are valid at all points. Tests against observations show that model-based initialization provides good guidance for coming weather.


Initialization is the provision of grids of initial surface weather elements, such as temperature, wind, cloud coverage, PoP, and so on, for the IFP editors.

Without initial grids of weather elements, a forecaster using the IFP system would be confronted with blank data display screens in the editors at the beginning of each forecast cycle. It is much easier for forecasters to modify a reasonable first guess of coming conditions than to create values from scratch with the editors. Initialization provides an automatic and reasonable set of future weather conditions which the forecaster modifies to represent the final best judgement of future conditions based on all information and experience.

IFP derives grids of initial surface weather values from many sources, including MOS, previous forecasts, and numerical atmosphere models. Future sources might include climatic averages and persistence. This position paper is an overview of generation of initial weather elements derived from numerical atmosphere models run at NCEP.

Numerical Weather Models used by Model-based Initialization

Presently we use the NCEP AVN, Eta, MesoEta, NGM and RUC models. We have used the FSL LAPS-RAMS combination running at FSL for the Colorado area, and plan to resume that service. Any good local model such as RAMS can be used for initial guidance.

Data Source used by Model-based Initialization

Initialization uses the same data files of model output as the WFO-Advanced (AWIPS) display system. The data is transmitted from NCEP via the Satellite Broadcast Network (SBN) and stored on WFO-Advanced as NetCDF files.

Some initialization development on RUC uses data not transmitted via the SBN to offices. RUC is in rapid development and we are testing the full capability of RUC, assuming that it will eventually be available in offices.

Weather Elements made by Model-based Initialization.

Presently model-based initialization of IFP makes the following weather elements at all grid points:

* These parameters are not well estimated by the current algorithms.

Time Periods covered by the Models, and Model Grid Resolution

Table 1 shows time periods covered by initialization from each kind of model, forecast time steps, and model grid resolution. IFP initialization nominally creates grids with 10 km resolution; that can be easily changed if needed.

Table 1: Model Time Coverage and Grid Resolution 
------------------------------------------------------------
Model    Time Coverage   Time Step  Model's Grid Resolution 
------------------------------------------------------------
AVN      0 - 72 hours     6 hours         80 km     
Eta      0 - 48 hours     6 hours         48 km      
MesoEta  0 - 33 hours     3 hours         29 km
NGM      0 - 48 hours     6 hours         80 km
RUC      0 - 12 hours     3 hours         60 km
------------------------------------------------------------

Geographic Areas covered by the Model-based Initialization

Model-based initialization can be run anywhere inside the model output grids provided by NCEP to the offices via SBN. Usually this includes the lower 48 U.S. states and adjacent waters and lands. The new Eta model (late summer 1997) will also cover Hawaii and Alaska, and model-based initialization will be available over all of those areas.

Derivation of Surface Weather Elements from Model Output.

Techniques for derivation and creation of weather elements from model output are described in IFP Position Paper 5. The code in use at the moment uses algorithms for clouds, precip phase type, and visibility from Jim Ramer, and ideas for estimating probability of thunder and severe weather from Keith Brill and Ed Szoke. The remainder of the code was developed by Stuart Wier.

Derivation of surface weather elements from model output provides these enhancements over raw model output for IFP:

Quality of Model-based Initialization

The values of many initial weather elements made from models are checked against observations at stations across the lower 48 states of the U.S. The quality of these initial forecasts is good. There are strong indications that IFP model-based initialization is ready for operational forecast tests.

Details of the tests and results are described in IFP Position Paper 4.

Advantages of Model-based Initialization

Model-based initialization takes full advantage of the high-quality models of the atmosphere now in use.

Model-based initialization instantly benefits from improvements made in the underlying models. Changes now happen rapidly in modeling, and IFP initialization benefits immediately from better model physics, spatial resolution, and time resolution, without re-coding or other delays.

Model-based initialization retains the full resolution of model output. The IFP system editors present the highly detailed model forecasts to the forecaster, and allow that high resolution detail to be preserved in forecast products.

Initialization on a high-resolution grid, based on a numerical model of the atmosphere, catches important weather developments in place, not only after they pass selected stations as is the case for station-based approaches. You can see the front coming before it reaches your location. You can see severe weather development between cities.

Having several models in initialization means that forecasters have a choice of which model to use.

Having several models in initialization means that the failure of a model or failure of arrival of some model data does not leave the forecaster empty-handed for 12 hours or more.

Graphical Displays of Model-based Initialization

Click on the selection to see a display:

Examples of today's initial forecast guidance from models are displayed by the AFPS Graphical Forecast Viewer (GFV) and Formatter available on the EFT home page. The GFV is in development and may not be available or have data at all times.

Future Work

Future work will include improvement in estimation of surface weather values from models, adaptation to new model features, and expansion into initial weather elements needed for aviation and marine forecasts, such as wave heights and superstructure icing.

One needed improvement is a good visibility algorithm. New output from Eta and RUC (cloud moisture amount, etc.) will allow trial of visibility algorithms based on actual optics an cloud physics. Good visibility will allow forecasting fog as well as visibility.

Another area of future work is orographic precip amounts, especially for snow amount. This is essential for good modeling of snow in the mountains. Since model-initialization already includes moisture, wind, and topography, no new data is required.

July 3, 1997; revised August 22, 1997.

Stuart Wier

NOAA, Forecast Systems Lab, Boulder, Colorado