Software development to support probabilistic forecasting on AWIPS II

1
Software development to support probabilistic
forecasting on AWIPS II

• 25th IIPS Conference
• Phoenix, Arizona

Paul Schultz, project manager
Tom LeFebvre, Tracy Hansen, Leigh
Cheatwood-Harris, Tom Kent, Jim Ramer, and Brian
Motta
14 January 2009
2
History, motivations, requirements

• NFUSE
• NRC report Completing the Forecast
• AMS Committee on Uncertainty Forecasting (ACUF)
• Hydrometeorological Testbed
• PQPF from mesoscale model ensemble
• ALPS workstations
• GSD investment
• Workstation development
• Forecast modeling and postprocessing
• Decision support

3
GSDs first efforts

• Focus on temperature, winds, and precipitation
• Not yet severe weather or aviation
• Focus on WFO problem
• But keep an eye on NC problem (NAWIPS)

4
Examples I
5
Examples II
6
Examples III
7
Examples IV
8
Workshop

• Held at GSD October 22-24
• Forecasters, program leads, developers
• Goals
• Initiate the discussion on opportunities and best
  practices
• Try out new workstation capabilities
• Give and get feedback
• Think about the forecast process

9
Forecast process, first guess

• Key components
• NWP
• Verification
• Climatology
• Obs
• SREF members (reforecasts)
• Analog finders
• Assumption forecasters can get to know a
  reasonable number of predictors (models), learn
  how they perform on certain local forecast
  problems
• Use AWIPS II to look for historical analogs
• Run verification stats on those cases
• Correct model output for biases
• Interactive member weighting
• By verification stats forecaster experience
• Edit PDFs

10
From left Paul Schultz, Scott Jacobs, Matt
Peroutka, Steve Amburn, Tracy Hansen, David
Metze, Larry Dunn, Rich Grumm, Tom LeFebvre,
Pablo Santos, Steve Schotz, and Jim Ramer. Not
shown Doug Hilderbrand, Leigh Cheatwood-Harris,
Asvin Mysore
11
Forecast process skeleton

• Key components
• NWP
• Verification
• Climatology
• Obs
• SREF member (reforecasts)
• Analog finders
• Assumption forecasters can get to know a
  reasonable number of predictors (models), learn
  how they perform on certain local forecast
  problems
• Use AWIPS II to look for historical analogs
• Run verification stats on those cases
• Correct model output for biases
• Interactive member weighting
• By verification stats forecaster experience
• Edit PDFs

12
Findings and Outcomes

• 1 Yes we can
• Not a given (NURO)
• Revised forecast process
• Todays NDFD temperature becomes the median (50)
• First guess for spread information comes from
  various sources
• Guidance EKDMOS, Downscaled NAEFS, NCEP
  reforecasts
• Amburn method for PQPF
• Based on PoP and QPF
• Krzysztofowicz Evans method for temps
• Fast-iteration project
• Software installed in small number of field
  offices
• On-line forum
• Bug fixes or new software in days not weeks
• Early advances towards AWIPS II
• Convergence of D2D and NAWIPS for ensemble
  manipulation
• Beta testing at HPC
• Opportunity to research SOA options for data
  distribution

13
Thank you

• Please visit our demo at the OAR booth
• Heres our web page

http//www.fsl.noaa.gov/ProbFcst
14
For the users

• Interactive web-based decision support tool
• Cost/loss oriented
• Experimental graphics
• Web, newspaper, etc.
• Text products

15
Sample text product
16
Sample graphic
17
EMP3
18
More EMP3