(back)


COMPOSING THE DISCUSSION

Up to this point you have been shown techniques to conduct both an initialization and prognostic assessment of the model.  The information gleaned from these techniques will afford you sufficient material to comprise the discussion.

The discussion is divided into three sections


The format of the discussion will help users isolate portions of the discussion pertinent to their interests.  The format also curtails superfluous detail.  Below is an example of a discussion...note how a particular meteorological feature of interest is carried in all sections as its own subsection.
 


Also note how a standard reference to the ADMNFD message is included at the beginning of the discussion AND a reference to the model biases are included at the end of the discussion.  This will allow the composer to minimize repetition of information in the discussion already available for public consumption.

In the initialization discussion, include the text .SEE NOUS42 KWNOADMNFD FOR STATUS OF UPR AIR INGEST. Discuss initialization errors only if they can be attributed to a notable impact on the models forecast. The initialization problem(s) with a specific feature will contain the same headline found in the .Model Trends. and .Model Differences and Preferences.. Otherwise, a general statement will be added after the .SEE NOUS42.. indicating no significant errors were noted in the (NAM/GFS) initialization. Examples . .NO SIGNIFICANT ERRORS NOTED IN THE GFS INITIALIZATION. or .NAM INITIALIZATION SEEMED REASONABLE.. An example of initialization problems would be when satellite imagery indicates development over the Pacific, which is not indicated in the models.

In the remainder of the discussion, be sure to describe where significant differences in model solution exist - and WHY you think these differences exist.  A good approach therefore for determining which features specifically to highlight in the discussion are those corresponding to expected significant weather. Specific quantification of differences and trends are not required. Relative or general terms are allowable. For example .THE NAM IS TRENDING FASTER WITH THE COLD FRONT ON DAY TWO. is satisfactory.

Although the PMDHMD content will focus mainly on mass fields, you should be aware of differences in QPF and precipitation type output in regions where significant weather is expected.  For example, look for problems in the model output relating to grid scale feedback from convection.

Since QPF and winter precipitation are covered in their own discussions, focus description of any differences to those impacting where forcing mechanisms or thermal profiles (i.e. - those that dictate how a forecaster should deviate from model guidance).

Once you complete the model differences and preferences section, coordinate your findings with those on shift with you before releasing the information to the public.  The discussion is built as NAM and GFS data arrives to HPC. There are two issuances per shift. The first issuance (0400Z/1600Z) includes the examination of the NAM while the second and final issuance (0530Z/1730Z) appends the examination of the GFS, model differences, and ultimately model preference as follows:

A discussion of the GFS and NAM must be contained in the discussion, however, other models may be included as well.

After each release, send a notification to participants on 12 Planet - and include a terse summary of the updated content.


VERIFICATION

What forecast would be worth composing without performing verification ?  Although there are no objective ways of quantifying your assessment, you can review the verification data to determine if your assessment lead users in the right direction.  Below is a two panel display of 500 mb heights and PMSL for the 72h period corresponding to the case above.  The Eta is in Red, the GFS is in Green and the verification (Eta analysis) is in blue.

As you can see the prognostic assessment was correct for the Pacific system at 500 mb (recall a GFS like solution was preferred).. particularly after fhr 60.  In the east its not so clear.  At about fhr 60 the Eta seems to have a better handle on what verified and by progression seems to have had a slight edge over the GFS at least in terms of location of the surface low.  And by fhr 72 at 500 mb the Eta was the most progressive of the two models and compared best to the verification - not by much however.

You may wonder that if a different result may have been offered if the verification was from the GFS analysis. Below is the same loop as above, except blue contours represent GFS analysis.  You might think if the verification is from the Eta, perhaps a comparison to a verification using the GFS analysis would offer a different result.  Below is the same as above, except the blue is the GFS analysis.

As you can see, the same signal is present for both the Pacific at 500 mb and even the surface assessment in the East.  The Eta advantage in the east is a bit more notable in this loop, but still not as dramatic as result in the Pacific.


FINISHED ?

This module has provided you with a foundation conducting short range model diagnostic assessments at HPC.  The information in this module can serve as a starting point for developing your model diagnostic expertise.  The following appendices specify some of the mechanics provided in this module.


APPENDIX A  (AWIPS Functionality useful for Model Diagnostics)

Sampling: Right click and hold the mouse over a satellite or gridded image to obtain a constantly updated display of value under the cursor

Convert AWIPS grid contours to an image: Right click and hold over an AWIPS product label at the bottom right of the screen - then select view as image.


APPENDIX B (Observational Data in AWIPS)

RAOB data can be obtained from D2D Menu by clicking UA PLOTS, then RAOB  (examples below are for 250 mb level):

Similarly, wind observations can be obtained from the same menu, but by selecting PIREP or MDCRS plots (and then the appropriate level):

PROFILER winds can be obtain for a level from the same menu, but by clicking PLOTS (and then the appropriate level):

Satellite winds are obtained by clicking the Satellite menu, then GOES High Density Winds, and then the appropriate layer:

PW observations can be obtained from the NCEP menu, then selecting Precip & Stability, and then Precip H20 Plot or Analysis:


APPENDIX C (Volume Browser)

To display specific output fields from a model, select Volume, and then BROWSER:

and a browser will display 

Choose your model under SOURCE, and the field of interest under FIELDS (Basic, Derived)..

("Derived" menu example)           ("Basic" Menu example)


APPENDIX D (AWIPS Procedures for Initialization Assessment)

All the above displays for initialization assessment are stored in a procedure on AWIPS.  To access this procedure, click
Procedures, then OPEN, then select "Models - Initialization NEW", then load the 5 procedures into the 5 panels on D2D


APPENDIX E (NAWIPS Prognostic Assessment Displays)

Verification loops can be found in NTRANS under the avnver or etaver selection:

 

Trend loops can also be found in NTRANS under a given model with the " _comp" listing:

Single time trends can be performed in NMAP by loading multiple model runs, but selecting SINGLE TIME FRAME on the time line window.  This will load each model valid at the single time you have specified on the time line.


APPENDIX F (List of referenced links)


Acknowledgments
Thanks to God for helping me with this and thanks to Jessica Clark (HPC Model Diagnostic Focal point) for helping review and refine the document.

Last updated 01/31/2005