This page links to graphics and other data from model experiments for the different branches of the EMC.  In no way do they represent operational forecasts, but we offer them so that meteorologists in the field may give us feedback on the results.  Diagnostic tools are used to analyze the date at EMC, and the results of this analysis are what appear as links below.

Before going to the data, however, we suggest you read the following section, which discusses the experimental process, from how EMC uncovers a problem to the final "parallel" run, performed to remove the final "bugs" prior to operational implementation.
 
 

How do we determine what experiments to perform to improved the numerical weather prediction (NWP) model?
 

• Through examination of existing model biases or error patterns, or
• Through research done at NCEP or elsewhere which might improve an existing NWP model, using
• new physical parameterizations,
• new data assimilation methods, or
• better resolution,
• Through anticipated increases in computer resources which drive decisions on what improvements can be implemented.

Terms typically used for model experiments before model changes become operational include, in roughly chronological order from problem identification to pre-implementation:

• case study (examine one weather event to diagnose model perfomance, for example)
• sensitivity study (low-resolution model runs to see if a model change will have a significant positive or negative impact on model climate)
• experimental runs (low-resolution forecasts without data assimilation, to check the changed model climate)
• retrospective (usually in conjunction with model climate testing)
• parallel (full data assimilation system at expected operational model resolution, with 'bundles' of experimental changes which have passed previous tests above)
• How do all the planned changes interact together?
• Assessment done by NCEP and field office Science and Operations Officers (SOOs)
• Forecast impact of:
• Amplification of changes and errors through the data assimilation system
• Changes in old biases and errors
• Effects on vertical motion (model spin-up)
• Accumulate vertification statistics for an objective comparison to existing operational model configuration

A schematic of the full model change cycle can be found here [link will be activated to flowchart "Path to Operations" soon!].
 

Global Modeling Branch experimental runs:

• Pete Caplan's near-realtime T170 parallel run diagnostics

• Suru Saha's web page

• Experiments to solve the GFS warm bias problem
 
Each of these experiments are 45 day model runs, with the first 15 days thrown away to account for model spin-up ("adjustment") to the model change.  Observations are NOT incorporated into the experiment except to determine the initial condition.  The experiments either involve the treatment of short and long-wave radiation in the atmosphere or the parameterization of surface processes like albedo or vegetation control over evapotranspiration.

Diagnostic comparisons of model bias are made to the NCEP Reanalysis product for both experiment and  operational GFS (the control).  Graphics of both time series of regional area averaged model bias, and regional spatial distribution of model bias are shown.
 

• T170 parallel run test diagnostics (summer  1998 retrospective and winter 1998-99 concurrent experiments)
• Climate parallel runs (climate version of the GFS run at coarse resolution)

• Yuejian Zhu's experimental web page  (Comparison of GFS output grid data to observations)

Includes comparisons from summer 1998 retrospective run, winter parallel runs for T170L42, and North American Observing System (NAOS) global experiments.
 
• Glenn White's systematic  differences and errors in the parallel test of T170/L42 in Feb. 1999

Also includes diagnostics for parallel test from August 1998 (for summer comparison in addition to winter comparison).  Zonal means and regional differences are displayed.
 
 
•  Glenn White's diagnostic comparison of old and new orography during February-March 1999

 
This is a low-resolution (T62) set of experiments using a new orography.  The effect of the new orography is most marked on gravity waves and topographically forced long waves.

Mesoscale Modeling Branch experimental runs:

Eta:

 
• Experimental Eta runs

• Parallel Eta-X.
• Parallel Eta runs at 48-km resolution, including Kain-Fritch convection scheme test, and assimilated precipitation from multisensor analyses.
• Kain-Fritch convective scheme
• NSSL runs
• Winter precipitation types (from Bob Rozumalski/SOOSAC, determined by postprocessing Eta model data)

 
 
 

Marine Modeling and Analysis Branch experimental runs:

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Climate Modeling Branch experimental runs: