Transcript TonyN.ppt

Ideas for NWS EPS
Advancement
Tony Eckel
Naval Postgraduate School
EPS Components
I.
Foundation: Observations, Data Assimilation, the
Model(s), Model Resolution, …
II.
Ensemble: Initial Condition Perturbations, Model
Perturbations, # of Members, …
III.
Exploitation: Post-processing, Products, Verification,
User Education, …
Design all components based on Value first, Skill second
• Evolving to “service paradigm” with focus on optimizing users’
decision processes requires different emphasis for measuring forecast
quality since skill ≠value
• Metrics of Value (e.g., ROCSS, VS, etc.) measure quality from
user perspective
• Metrics of Skill (e.g., BSS, CRPS, etc.) are important for scientific
evaluation
• Requires intimate relationship with users to understand their weather
sensitivities and risk tolerances
Post-Processing
• Truth: Must capture phenomena and scales of concern to user
• Reforecast dataset required for robust calibration
– Match EPS design (no short cuts!)
– Length dependent upon capturing user phenomena
• Down-scaling critical to value
• Meteorological consistency within each member may be
challenging issue
Thorough and Open
Verification
• Critical to building user confidence
• Focused on user sensitivities
• User-friendly: web-based, interactive, well documented, etc.
• Continuously updated
• Link with products and education
User Education
• Need campaign (concerted effort) to teach users methods for
optimal decision making given forecast uncertainty information
• Broad-based: Specific users and general public
• Include detailed strengths and weaknesses of the EPS