Lessons learned and initial best practices converting from mdl matrices to gfe formatters
Download
1 / 35

Lessons Learned and Initial Best Practices Converting from MDL matrices to GFE Formatters - PowerPoint PPT Presentation


  • 69 Views
  • Uploaded on

Lessons Learned and Initial Best Practices Converting from MDL matrices to GFE Formatters. David Zaff NOAA/NWS Weather Forecast Office, Albany, New York November 4, 2003. ALB IFPS Methodology Before (prior to 06/03). 1 st draft: GFE 2 nd draft: Matrix

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about ' Lessons Learned and Initial Best Practices Converting from MDL matrices to GFE Formatters' - lola


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
Lessons learned and initial best practices converting from mdl matrices to gfe formatters

Lessons Learned and Initial Best Practices Converting from MDL matrices to GFE Formatters

David ZaffNOAA/NWS Weather Forecast Office, Albany, New YorkNovember 4, 2003


Alb ifps methodology before prior to 06 03
ALB IFPS Methodology Before(prior to 06/03)

  • 1st draft: GFE

  • 2nd draft: Matrix

    • All fine tuning done through matrix

    • All updates done through matrix – Grids would not change for up to 12 hrs.

  • 3rd draft: post editing ZFP

  • 4th draft: post editing SAF


Alb ifps methodology before prior to 06 031
ALB IFPS Methodology Before(prior to 06/03)

  • More time was often spent on the matrix than on the grids

  • Little detail in grids

  • Grids showed poor consistency

    • Between offices

      • (no use of ISC)

    • Meteorologically across our grids

      • (ex Def R- w/ a 50% PoP)

    • Forecast vs. Reality

      • Both in short term and meteorologically (unrealistic temp changes)


Alb ifps methodology now
ALB IFPS Methodology NOW

  • GFE always open

    • GRIDS are the “flagship product”

  • ZFP/SAF produced from the GRIDS

    • (Actually all text products are produced from grids)


Alb ifps methodology now1
ALB IFPS Methodology NOW

  • Biggest change:

    • We’ve taken a step out of the forecast process

    • We have a better relationship between the grids and the text


Lessons learned
Lessons Learned

  • 1) Any Workload improvement?

  • 2) Matrix vs GFE Formatters – any time savings?

  • 3) Four Reasons the text is post edited

  • 4) Amount of text editing varies from forecaster to forecaster

  • 5) The SAF and ZFP


Lessons learned1
Lessons Learned

  • 1) Is there any workload improvement??


Lessons learned2
Lessons Learned

  • 1) Is there any workload improvement??

  • YES!!!

  • Because you remove a step in the forecast process – there is indeed a workload improvement

  • BUT – you spend more time and effort in the grids

  • Don’t forget – this is where the meteorology is!

“grid ache”


Lessons learned3
Lessons Learned

  • 2) Is producing text quicker via GFE Formatters than with the Matrix?


Lessons learned4
Lessons Learned

  • 2) Is producing text quicker via GFE Formatters than with the Matrix?

  • Yes and No!

    • GFE Formatters are a HUGE time saver

    • GFE Formatters do not save any time because you have to put much more effort into the grids

  • The point is that while in the grids, you spend more time thinking in terms of meteorology and less time working with text output


Lessons learned5
Lessons Learned

  • 3) Four Reasons the text is post edited (in the post Matrix era):

    • 1) Grid problems

    • 2) Large zone groupings (we use Xnow)

    • 3) Cosmetic wording changes

    • 4) Formatter problems


Lessons learned6
Lessons Learned

  • Four Reasons the text is post edited:

    • 1) Grid Problems:

      • Most text issues are grid related

        • Often the grids do not match the desired text

      • Problem 1:

        • It is easy to modify the grids to “fit” the desired text. The result is that the grids are not consistent in reality or meteorologically

        • There is a learning curve here! The ability to adjust grids to match text w/out compromising grids takes time


Lessons learned7
Lessons Learned

  • Four Reasons the text is post edited:

    • 1) Grid Problems:

      • Most text issues are grid related

        • Often the grids do not match the desired text

      • Problem 2:

        • Often it is too easy to simply edit the text.

          • It takes time to adjust the grids and rerun the formatter

        • This leaves the problem for the next forecaster


Lessons learned8
Lessons Learned

  • Four Reasons the text is post edited:

    • 2) Large zone groupings

      • We use Xnow – there are other alternatives

      • Problem:

        • The zone groupings are determined independent of GRIDS, and output therefore doesn’t always turn out as expected.

      • Solution 1: Regroup zones

      • Solution 2: Nudge grids (not recommended in most cases)


Lessons learned9
Lessons Learned

  • Four Reasons the text is post edited:

    • 2) Large zone groupings

      • We use Xnow – there are other alternatives

      • Ultimate solution: One zone/county one group!

        • While this will increase the size of the ZFP and therefore increase proofreading time, it will greatly improve formatter’s capability – less errors overall!


Lessons learned10
Lessons Learned

  • Four Reasons the text is post edited:

    • 3) Cosmetic wording changes

      • Forecasters need to let go of wording that do not meteorologically add to the forecast.

      • Example:

        • GFE output - Partly cloudy this morning then becoming mostly cloudy

        • Modified text - Increasing clouds

      • While the above change makes for a cleaner forecast, it does not add meteorology to the forecast.


Lessons learned11
Lessons Learned

  • Four Reasons the text is post edited:

    • 3) Cosmetic wording changes

      • Cosmetic wording changes are not an efficient use of forecaster time.

      • Our expertise and therefore our time should be put into the grids.


Lessons learned12
Lessons Learned

  • Four Reasons the text is post edited:

    • 4) Formatter Problems:

      • Pros:

        • You have total control

      • Cons:

        • It takes time to fix problems without overriding grids or introducing new errors


Lessons learned13
Lessons Learned

  • 4) Amount of text editing varies from forecaster to forecaster

    • Tough problem

      • The GFE has a very steep learning curve

        • (and you thought differential equations were hard!!)

      • Some take to the grids like a video game, and pick up smart tools and and ways to make the “grids work” such that the text comes out as expected.


Lessons learned14
Lessons Learned

  • 5) The SAF and ZFP

    • This is ALY’s biggest problem:

      • We massage the grids for the ZFP, with (hopefully) minimal text edits.

      • THEN, we have do it all over again with the SAF


Lessons learned15
Lessons Learned

  • 5) The SAF and ZFP

    • The SAF contains large sampling areas

    • Larger areas mean more potential wording issues

    • Same problem with large zone groupings


Lessons learned16
Lessons Learned

  • 5) The SAF and ZFP

  • Solution??

  • We don’t have one yet!

    • An easy one would be to ABANDON THE SAF, and simply use the ZFP

    • Better – add in local effects.


Initial best practices
Initial Best Practices

  • Quickly test wording by looking at one zone group and one period output.

  • Control text detail without modifying grids

  • Test/Non-operational GFE box

  • Get everyone involved!

  • Issuance Times


Lessons learned17
Lessons Learned

  • One more comment:

  • ALY switched from Matrix to GFE formatters in June, AND we switched from NWS directive C-11 to 10-503

  • How many folks contacted us about the change???


Lessons learned18
Lessons Learned

  • One more comment:

  • ALY switched from Matrix to GFE formatters in June, AND we switched from NWS directive C-11 to 10-503

  • How many folks contacted us about the change???

    • ZERO!!!!


Initial best practices1
Initial Best Practices

  • Quickly test wording by looking at one zone group and one period output.

    • It normally takes 20s-30s to run the formatter for all zone groupings and all forecast periods To make a change to the text you:

      • edit the grids

      • rerun the formatter

      • wait ANOTHER 20-30 seconds…

    • This process begs the forecaster to ignore the grids and focus on the text, introducing grid-text discrepancies!

    • The formatter runs VERY quickly when running through one zone, one period!!!!!!!!


Initial best practices2
Initial Best Practices

  • Quickly test wording by looking at one zone group and one period output.

    • To run, you need to choose three items:

      • Choose issuance time

      • Choose period of interest

      • Choose zone grouping (ex NYZ052>054)

    • Then you can rerun the above over and over again easily after quick grid edits.


Initial best practices3
Initial Best Practices

  • Control Detail text detail without modifying grids

    • GFE Formatters come with a default set of statistical sampling techniques can be overridden semi-permanently, but these cannot easily be changed on the fly.

    • The MDL matrix allows the forecaster to change the wording detail

    • ALY formatters preserve this useful capability via a GUI designed by Vasil Koleci and local formatter adjustments.


Initial best practices4
Initial Best Practices

  • Control Detail text detail without modifying grids

    • With this tool, you can have a high amount of detail in the grids, and gloss over some of this in the text, or leave ALL the detail if you prefer.

    • Problem: forecasters reluctant to simplify text, and high detail in wording can be confusing.


Initial best practices5
Initial Best Practices

  • Test/Non-operational GFE box

    • This is ABSOLUTELY CRITICAL to testing nearly everything related to the GFE:

      • Smart Tools

      • Smart Initialization

      • And most importantly – FORMATTERS

    • You are SEVERELY limited and will place yourself in a dangerous spot if you work on formatters/GRIDS that are operational.


Initial best practices6
Initial Best Practices

  • Test/Non-operational GFE box

    • The GFE provides the software needed to quickly transfer GRIDS, Formatters, Smart tools, etc from one machine to another.

    • With an RPP box, you can work on operational grids and formatters in realtime….but offline


Initial best practices7
Initial Best Practices

  • Get everyone involved!

    • The grids are an ever changing, always current database.

    • Note – ALY is not doing this…


Initial best practices8
Initial Best Practices

  • Get everyone involved!

    • Why write nowcasts saying it’s going to rain in 2 hrs, when the grids only have a chance of rain?

  • Springfield MO: the whole staff is involved

    • HMTs, Interns, all levels of forecasters and the management team


Initial best practices9
Initial Best Practices

  • Get everyone involved!

    • One person on short term

    • One person on next significant Wx event

    • One person on extended

    • Leave all other grids as is to preserve consistency and prevent “yo-yo-ing” forecast

      • (Use MOS data to update Max/MinT if you find this improves point-based verification stats)


Initial best practices10
Initial Best Practices

  • Issuance Times

    • If you have an “always up to date set of grids”, then why have issuance times?

    • You ought to be able to send out a new text product once an hour…


ad