National Blend of
Global Models
An Overview
Jeff Waldstreicher
NOAA/NWS Eastern Region HQ
Bohemia, NY
15th Northeast Regional Operational Workshop
November 12-13, 2014 – Albany, NY
Origins of the National Model Blend
REASONING
Local / Regional
Model Blend Initiatives
WRN Roadmap
NAPA Report
FUNDING
Sandy Supplemental
JPSS Gap Mitigation:
projects that make better
use of existing model data
2
Early Roots in Regional Blends
• Improvements in verification and consistency scores
for the NWS National Digital Forecast Database
(NDFD)
• Operational Efficiencies
BUT…
• Sustainability of the regional solution?
• Consistency between regional approaches and
national centers?
• Model data dissemination limitations?
Early Roots in Regional Blends
Need a national system:
– Harness full capabilities of NWP systems and
national computing power
– Full suite of ensembles
– Employ state of the art statistical post processing
techniques
• Downscaling
• Bias Correction
– Enable enhanced consistency within NWS forecast
information
National Blend of Models
National Model Blender Project
Develop a set of nextgeneration foundational
gridded guidance products
for NDFD weather
elements based on NWS
and non-NWS model
information.
NDFD
5
National Model Blender Project Objective
Improve quality and consistency of the
NWS National Digital Forecast Database
Improve operational efficiency
6-8 hrs. of manual
grid manipulation
Current NWS Forecaster
Leverage skill of guidance to assist with freeing up time
to focus on DSS and hazardous weather operations
6
NDFD Elements
NDFD Elements
No. of Grids
12-hour Probability of Precipitation (PoP12) 13-15
Apparent Temperature
33-41
Dew Point
33-41
Hazards
59-82
Maximum Temperature
Minimum Temperature
Quantitative Precipitation Amount
Relative Humidity
7-8
6-7
9-13
33-41
Significant Wave Height
Sky Cover
17-21
33-41
Snow Amount
Temperature
5-9
33-41
Weather
33-41
Wind Direction
33-41
Wind Gust
Wind Speed
17-25
33-41
Projections
Every 12 hours, out to 168 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every hour, out to 72 hours;
every 6 hours out to 120 hours
Every 24 hours, out to 168 hours
Every 24 hours, out to 168 hours
Every 6 hours out to 72 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every 6 hours out to 120 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every 6 hours out to 48 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
Every 3 hours out to 72 hours
Every 3 hours out to 72 hours;
every 6 hours out to 168 hours
7
The National Blend Project
Create a NWS methodology
for a national blend (“best”)
product, beginning with the
Day 3 - 8 time frame (Phase I)
Extensible to a full set of NWS
deterministic and probabilistic
products covering days 1-10
(Phase II).
8
National Blend Project – Phase I
Focus on Days 3- 8, so initial development efforts are
concentrated on global deterministic and ensemble systems
Gridded Guidance Products:
•
•
•
•
•
•
•
Max/Min Temperature
3-hr Temp and Dewpoint
Wind speed and direction
Sky Cover
PoP
Precipitation Type
QPF
9
National Blend Project – Phase II
Additional Elements:
•
•
•
•
Snowfall Amount
Wind Gusts
Predominant Weather
Others as needed*
*Derive where it makes sense to ensure
consistency, efficiency: i.e. Relative
Humidity, Apparent Temperature
Expansion to Days 1-10
• Non-global NWP systems
(e.g., NAM, SREF)
Development of a suite of
probabilistic guidance
10
The National Blend Plan
The National Blend will be distributed
twice daily.
WPC
Weather
Forecast
Offices
WPC will provide oversight of the
National Blend for Days 3-8 to maintain
meteorological and spatial consistency.
WFO’s will receive both objective and
WPC (edited) grids.
11
8 Sub-Projects Funded from Sandy Supplemental
~$2.2 million
• Lead efforts of post-processing team, contribute GEFS
Reforecast data sets, contribute to PoP/QPF development and
new blending approaches, participate in evaluation (Hamill;
OAR/ESRL/PSD)
• Advanced MOS statistical post-processing techniques for
calibration of NDFD variables and blending of model output
(Gilbert; OST/MDL)
• Expand NAEFS post-processing (Zhu; NCEP/EMC)
12
Sub-Projects Funded from Sandy Supplemental
~$2.2 million
• Accelerate URMA/RTMA improvements for use as “Analysis of
Record” in model blender (DiMego/Pondeca; NCEP/EMC)
• Perform grid-to-grid and grid-to-point verifications (Ruth;
OST/MDL)
• Data management and visualization tools to improve
collaboration (Gilbert; OST/MDL)
• Transition NCEP’s WPC blending and ensemble visualization
capability to AWIPS II for the NCEP Perspective (Mainelli;
NCEP/NCO)
• Training (Mostek/Bua; COMET)
13
Consistency - NWS Standard Terrain
Different terrain used in downscaling systems
(AWIPS D2D display, GFE, RTMA)
GFE - RTMA Terrain
All shaded areas represent
differences greater than 100m
Example: Gridded MOS 2-m T @ 12 hours,
diff > 10oF
14
Consistency - NWS Standard Terrain
• Development groups, including those responsible for
post-processing and those responsible for the analysis
will downscale to the same terrain in use by the WFOs
in GFE.
Develop methodology with schedule to merge local edits
into a national terrain file
Develop procedures to create a gridded land/water mask
for MDL and EMC downscaling applications consistent with
the AWIPS terrain file and marine polygons
Scheduled for AWIPS 14.4.1, baselined by June 2015
Remaining Question: How do local terrain edits get back to
a national file?
15
Project Management
Project
Advisor
Steve Lord
Project
Manager
Kathy Gilbert
Plenary Group (Regions,
NWSEO, NCEP, OST, OCWWS,
SPPO, OAR)
Working level
Teams
Analysis and
Verification
(Ruth)
Post
Processing
(Hamill)
Testing and
Training
(Novak and
Craven)
Dissemination
(Settelmaier
and Gilbert
Outreach
(Sieveking and
Gagan)
National Blend – Analysis & Verification
Multi-Panel Graphical Interface created by MDL:
http://www.mdl.nws.noaa.gov/~blend/blender.prototype.php
•
•
•
•
•
Initial evaluations of individual components of the blend
Prototype National Blend grids
Provide feedback to RTMA developers
Provide feedback to Blend developers
Show verification of experimental blends
Coming Soon:
• Ability to overlay surface observations
• Grid-based and point-based verification statistics for
prototype blends
17
RTMA/URMA Evaluation
Blend Evaluation
National Blend Project - Evaluation
• Internal Evaluation through Summer 2015
(Testing and Evaluation Team)
• Formal Evaluation Sept 2015 – Mar 2016
• Analysis of Accuracy, Consistency, and
Workload (focused on NDFD, WPC, and
National Blend)
WFO input will be key
component of evaluation.
20
National Blend Project - Training
•
•
•
•
•
•
Overview of the National Blend
Components of the National Blend
Statistical Techniques & Procedures
National Blend Verification
National Blend Reliability
WPC Procedures
21
Questions?
…or time for some liquid blends…