1 agenda topic: national blend presented by: kathryn gilbert (nws/ncep) team leads: dave myrick,...

6
1 Agenda Topic: National Blend Presented By: Kathryn Gilbert (NWS/NCEP) Team Leads: Dave Myrick, David Ruth (NWS/OSTI/MDL), Dave Novak (NCEP/WPC), Jeff Craven, Jim Sieveking, John Gagan (NWS/CR), Tom Hamill (OAR/ESRL/PSD), Jack Settelmaier (NWS/SR) Contributors: NWS Regions, NWS HQ, NCEP Centers, OAR/ESRL

Upload: solomon-wilkins

Post on 03-Jan-2016

213 views

Category:

Documents


0 download

TRANSCRIPT

1

Agenda Topic: National Blend

Presented By: Kathryn Gilbert (NWS/NCEP)

Team Leads: Dave Myrick, David Ruth (NWS/OSTI/MDL), Dave Novak (NCEP/WPC), Jeff Craven, Jim Sieveking, John Gagan

(NWS/CR), Tom Hamill (OAR/ESRL/PSD), Jack Settelmaier (NWS/SR)

Contributors: NWS Regions, NWS HQ, NCEP Centers, OAR/ESRL

The Growing Challenge: Consistency

Issue: Local forecast offices work primarily to serve local user requirements. However, are state, regional, and national needs being addressed?

Consequence: Inconsistencies across CWA or Regional lines can lead to challenges for IDSS on the state, regional & national scale.

• What do our partners think when they see sharp changes in our forecast grids? - It impacts their confidence in our forecasts- They remove NWS products from their briefings

Need: To develop nationally consistent methodologies

2

3

Operational System Attribute(s)First implementation December 2015

System Name Acronym Areal Coverage Horz Res

Cycle Freq

Fcst Length

(hr)

National Blend of global Models - CONUS

NBM CONUS NDFD expanded domain

2.5 km 2/day 192

NBM - Oceanic NDFD Oceanic 10 km 2/day 192

NBM - Alaska NDFD Alaska 3 km 2/day 192

NBM - Hawaii, Puerto Rico NDFD HI, PR 2.5 km 2/day 192

System Attributes

MOS Calibrated GFS, linear regression models

EKDMOS Calibrated GEFS and CMCE, linear regression models

GFS, GEFS Bias-corrected GFS and GEFS direct model inputs

CMC Ensembles Bias-corrected, direct model inputs (GEM)

RTMA/URMA Analysis for bias correction and verification, future calibration

System Data Assimilation or Initialization Technique

4

Why System(s) are Operational Primary stakeholders and requirement drivers

• Stakeholders: NWS field offices, NCEP centers, all users of NDFD• Drivers: NAPA Report, Disaster Relief Appropriations Act of 2013

What products are the models contributing to?• National Digital Forecast Database

What product aspects are you trying to improve with your development plans?• Consistency of products through a common starting point, improve

collaboration between offices• Increased Decision Support Services• Quantifying uncertainty through probabilistic output from ensembles • Increased accuracy

Top System Performance Strengths• Runs on the same supercomputer as the operational models• Spatially consistent• Extensible and maintainable• Project will downscale model data, analyses and forecasts to a common

elevation/terrain datasetTop System Performance Challenges

• Never enough disk space to easily process ensemble members• Need to be as good or better than the current regional blending techniques• Complex project, not known yet how to optimally blend all fields

5

System Evolution Over the Next 5 Years

Major forcing factors• NAPA Report – “Forecast for the Future: Assuring the Capacity of the Nationa

l Weather Service” Recommendation: Improve consistency of products and services across the organization

• Support the goals of the Weather Ready Nation, i.e. Improve Weather Decision Services

Science and development priorities• Complete development of all NDFD weather elements (days 3 – 8) for all

NDFD CONUS and OCONUS domains in Phase 1 • Add mesoscale models and aviation weather elements to improve days 1-3• Develop probabilistic products from calibrated ensembles to quantify

uncertaintyWhat are your top challenges to evolving the system(s) to meet

stakeholder requirements?• High expectations, i.e. timelines, maturity of techniques• Skill of current GFS/GEFS inputs less than skill of comparable ECMWF fields• Requires large amounts of high-quality training datasets and disk space

Potential opportunities for simplification going forward• Subversion repositories accessible from R&D machines to share common

libraries with potential contributors• Rely on free and accessible data

6

Top 3 4 Things You Need From the UMAC

1. Advocate for retrospective runs before the models are upgraded• Build into timelines so calibrated model output is ready for

implementation with the model upgrades

2. Commitment to an Analysis of Record of sufficient length and quality for calibration, bias correction and verification• For all weather elements in the National Digital Forecast

Database and all forecast domains, including the Oceanic domain

3. Evaluate NOAA’s ability to meet its future requirements in a way that includes post-processing• Can requirements be met through a combination of less

computationally expensive post-processing and models?

4. Commitment to make America’s NWP the best global model in the world with full and unrestricted access.