The New National Weather Service: Building a Weather-Ready Nation, NCEP Priorities and Operational Modeling in NOAA Dr. William M. Lapenta Director, National Centers for Environmental Prediction NOAA/National Weather Service 2015 Weather and Climate Summit 12 January 2015
National Weather Service 2 Presentation Outline Weather Ready Nation & the National Centers for Environmental Prediction Production Suite and the Supercomputer Numerical Weather Prediction 101 Model Development Process 101 Ensembles Hurricane Sandy October 2012 and Impact on Modeling Thoughts Going Forward into 2025
My Intent Today.. Provide you with some unique insight and the opportunity to ask questions about NOAA operational modeling. 20 years at NASA performing research *** 06 years at NOAA managing operations ***My basic understanding of operational modeling at NCEP was quite poor National Weather Service 3
NWS Strategic Outcome: A Weather-Ready Nation Becoming a Weather-Ready Nation is about building community resiliency in the face of increasing vulnerability to extreme weather Ready, Responsive, Resilient REQUIRES NWS TO: Fully Integrate our Field Structure: Better Forecasts and Warnings Ensure Consistent Products and Services Provide Impact-based Decision Support Services (IDSS) Deliver through Multiple and Reliable Dissemination Pathways Work with Partners to gain needed response; includes embedding NWS in Emergency Operations Centers National Weather Service Involves entire US Weather Enterprise WORKING TOGETHER to achieve far-reaching national preparedness for weather events 4
IDSS & Communication Buffalo NY Lake Effect Snow: Nov 19-21, 2014 NWS Buffalo Facebook National Weather Service 5 NWS Buffalo 36 hrs prior to event NY Post
NWS National Centers for Environmental Prediction Specialized Services Common Mission 9 National Centers with ~650 Employees Aviation Weather Center Climate Prediction Center Environmental Modeling Center National Hurricane Center NCEP Central Operations Ocean Prediction Center Space Weather Prediction Center Storm Prediction Center Weather Prediction Center Vision: The trusted source for environmental predictions from the sun to the sea, when it matters most Mission: Deliver national and global operational weather, water and climate products and services essential to protecting life, property and economic well-being National Weather Service 6
What NCEP Delivers Provision of Services from the Sun to the Sea Solar Monitoring, Warnings and Forecasts Climate Seasonal Forecasts El Nino La Nina Forecast Weather Forecasts to Day 7 Extreme Events (Hurricanes, Severe Weather, Snowstorms, Fire Weather) Aviation Forecasts and Warnings High Seas Forecasts and Warnings Model Development, Implementation and Applications for Global and Regional Weather, Climate, Oceans and Space Weather International Partnerships NOAA Research to Operations Testbeds Super Computer, Workstation and IT Network Operations 7
National Weather Service 8 Strategic Areas for NCEP in the next 5 Years A major player in the Weather Ready Nation Strategically expand science/service areas based on user requirements: Extending lead time for high impact events Incorporate a full earth system science approach 3-4 weeks forecasts (closing the gap between weather and climate) Strategically transition research into operations Deliver (with partners) the WRN integrated field structure Deliver world class operational numerical guidance required to support the WRN Deliver timely, reliable and accurate products and services Deliver high capacity IT infrastructure support, high performance computing and technical management services
NWS Strategic Goals Improve Weather Impact-Based Decision Support Services Improve Water Forecasting Services Enhance Climate Services and adapt to climaterelated risks Improve sector-relevant information in support of economic productivity Enable environmental forecast services supporting healthy communities and ecosystems Sustain a highly skilled, professional workforce equipped with training, tools, and infrastructure to meet mission National Weather Service Relationship Between a Weather- Ready Nation and Modeling Operational numerical guidance: Foundational tools to used to improve public safety, quality of life and make business decisions that drive U.S. economic growth Prediction is what makes NOAA/NWS unique and indispensable! 9
Operational Numerical Prediction Enterprise Global Observations ~2 billion/day 99.9% remotely sensed, mostly from satellites Data Assimilation and Models 3D-EnVAR 4D-EnVar (Early 2016) Earth System model; coupled Global resolution (27km)--(13km Jan 14, 2014) North America @ 12km and 4x/day CONUS: HRRR @ 3km 24x/day On demand fire weather @ 1.3km Statistical Post Processing & Reforecasts NOAA Operational Computer Transition to new IBM idataplex in 2013 208 trillion calc/sec ~ 3X increase to 700 trillion calc/sec in January 2015 Basis for a Seamless Suite of Climate -> Short Term Weather Prediction National Weather Service 10
1955 - US Weather Bureau First Operational Production Suite Baroclinic: 300 km / 3-layer CONUS to 36 hr 1500z run Started 2100z Finished 2230z Thermotropic: 375 km / 2 layer No. Am to 36 hr 0300z run Started 1000z Finished 1230z 500mb Height 72h Forecast Valid 03Z 11 March 1956 Barotropic: 600 km / 1 layer ~NHemis to 72 hr 0300z run Started 1230z Finished 1300z 11
Seamless Suite of Operational Numerical Guidance Systems 12
NOAA s Operational Numerical Guidance Suite (December 2014) 13
14 NOAA Weather and Climate Operational Supercomputer System (WCOSS) 24-h Cycle 29 December 2014 Number of Nodes GFS SREF RTOFS NAM HiRESW Waves GEFS HRRR 00 06 12 18 00 Time of Day (UTC) CFS
Computational Cost of Production Suite Components 15 % utilization based on the number of nodes used in a 24 hour period Acronym Component HRRR High Resolution Rapid Refresh NAM North American Model SREF Short Range Ensemble Forecast CFS Climate Forecast System GEFS Global Ensemble Forecast System HiRESW High Resolution Windows GFS Global Forecast System MAG Model & Analysis Graphics WAV WAVEWATCH III prodser COMMS Overhead RTOFS Real Time Ocean Forecast System RAP Rapid Refresh GWVENS Global Wave Ensemble RDAS Reagional Data Assimilation System GDAS Global Data Assimilation System FVS Forecast Verification System Regional Systems 45% of total
Modeling CONUS & O-CONUS: Expensive and Requires Nesting 12% 3% 5% 22% North American Model (NAM) Runs 4x/day Outer grid at 12 km to 84hr Multiple Nests Run to ~48hr 4 km CONUS nest 6 km Alaska nest 3 km HI & PR nests 1.3km DHS/FireWeather/IMET Rapid Refresh (RAP) Runs 24x/day to 18h 13 km Domain includes Alaska 3 km HRRR runs 24x/day out to 15h High Res Windows (HiRESW) Typically run 2x/day 4 km WRF-NMM 5.15 km WRF-ARW 48 hr fcsts from both Short Range Ensemble Forecast System run 4x/day 21 members out to 72h 16km CONUS grids
NWP Building Blocks: Process Starts with Observations 17
What is Data Assimilation? The process of combining prior knowledge of the state of the atmosphere with new observations Prior knowledge = model forecast GOAL: Create an accurate representation of the atmosphere at time T=0 of the forecast process (Analysis) Methods: Optimal Interpolation Variational (3D and 4D) Ensemble Based Hybrid (En-VAR)
Representation of Physical Processes 19
Importance of Model Resolution Example--Low (Coarse) Resolution 20 4 cells cover convection X X X
Increasing (high) Grid Resolution 50 cells cover convection Therefore, able to resolve feature 21
GFS 500-hPa Height Anomaly Correlation Frequency Distribution (1996-2013) 00Z Cycle GFS Day-5 Forecast Northern Hemisphere 0.7 0.9 Frequency (%) Frequency of poor forecasts (AC < 0.7) decrease Frequency of good forecasts (AC > 0.9) increase Anomaly Correlation 22
GFS Skill Improvement Due to Resolution, Data Assimilation and Physics Upgrades Percentage of GFS 5-Day 500mb Anomaly Correlation Greater Than 0.9 Percentage of Good Forecasts Northern Hemisphere Southern Hemisphere 38km to 27km New shallow convection; updated SAS and PBL; positive-definite tracer transport Flow-dependent error covariance; Variational QC 55km to 38km OSU 2-L LSM to 4-L NOHA LSM 70km to 55km Hybrid-Ensemble 3D-VAR Data Assimilation AMSU-A & HIRS-3 data Year 23
Development and Transition to Operations Process I HAD NO CLUE 24
2014 HWRF Development Test Plan Infrastructure/DA Upgrades Physics upgrades Combined (best of all upgrades) H14A H14B Nest motion (H140) NOAH LSM (H141) Upgraded Ferrier (H142) RRTMG (H143) Ocean (H144) H214 (proposed 2014 HWRF) Description 1. Sat Da with 61 levels, 2mb top 2. Extended d02/d03 3. Upgraded vortex init. 4. GSI upgrades 1. No Sat DA 2. Include Invests in cycling 1. New nest motion & diagnostics NOAH LSM Separate species and F_rime advection with other upgrades Radiation MPI-POM with new coupler Baseline + physics + python based scripts *need to do test runs with new GFS in WCOSS Cases Whole 2011-2013 & selected 2008-2010 storms Whole 2011-2013 & selected 2008-2010 storms Priority cases Priority cases Priority cases Priority cases Priority cases Whole 2011-2013 & selected 2008-2010 storms Due date Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 Feb. 15 March 31 Platform Jet/WCOSS Jet/WCOSS Jet Jet Jet Jet Jet Jet/WCOSS 25
National Weather Service 26 2014: Improved Forecasts & Products Hurricane WRF Model & New Storm Surge Graphics HWRF Forecast Track & Forecast Radar Reflectivity Hurricane Arthur: July 01-07, 2014 18z 01 JUL 14 Potential Storm Surge Mapping Best Guess, Worst Case Scenario 16z 02 JUL 14 Actual Track HWRF Forecast Track Track Errors (nm)* 24hr 48hr 72hr HWRF ~25 ~45 ~120 GFS ~25 ~55 ~120 ECMWF ~45 ~130 ~310 *preliminary numbers
Ensemble Prediction NMC 1992: First operational global ensemble forecast system (GEFS) Global spectral, 3 members, ~210km / 18 layers to 240 hours NCEP: 2001: Short range Ensemble Forecast (SREF) 10 members @ 48 km, North America out to 63 hours 27
Ensemble Spread (Track) Provides Measure of Uncertainty 28
29 Ensemble Postage Stamp Plots Provides Measure of Uncertainty Storm location at day 5 for 21 ensemble members
Hurricane Sandy October 21 through October 31, 2012 National Hurricane Center's 5 day forecast track issued at 11 a.m. EDT on Thursday, October 25 30
Raises Public Awareness of Modeling http://www.nbcnews.com/video/nightly-news/51108647#51108647 http://www.washingtonpost.com/blogs/capital-weather-gang/post/to-be-the-best-in-weather-forecasting-why-europe-is-beating-the-us/2013/03/08/429bfcd0-8806-11e2-9d71-31
ECMWF vs NCEP Global Models Legacy Verification Metric (1984-2013) 32 Anomaly Correlation 0.05 AC Points
33 NOAA Operational Computer Upgrade in 2015 Petaflop = 1000 Trillion Floating Point Operations per Second 1 trillion = 1x10 12 = 1,000,000,000,000 NOAA Today= 0.27Pf Feb 2015= 0.776Pf Dec 2015= 2.5Pf Capacity per cluster ECMWF 2012= 1.5 Pf Today= 3.6Pf
Major Model Upgrades Planned for 2015 Acronym Component FY15 FY16 Q2 Q3 Q4 Q1 CMAQ Air Quality (Ozone and Smoke) X X GDAS/GFS Global Forecast System X X GEFS Global Ensemble Forecast System X GWVENS Global Wave Ensemble X HiRESW High Resolution Windows X HRRR High Resolution Rapid Refresh X HYSPLIT Smoke and Ash X MAG Model & Analysis Graphics X X X X NAM North American Model X RAP Rapid Refresh X RTMA Real Time Mesoscale Analysis X X RTOFS Real Time Ocean Forecast System X X SREF Short Range Ensemble Forecast X WAV WAVEWATCH III X GFS and GEFS: increasing resolution, data assimilation, physics, dynamics RAP & HRRR: Data assimilation and physics improvements SREF: Adding 6 members (27 total) and dispersion improvements, remove NMME d-core NAM: 3km CONUS nest, data assimilation, physics 34
Global Data Assimilation and Forecast System Scheduled Implementation is 14 January 2015 Global Spectral Model Semi-Lagrangian formulation 13km resolution from 0 to 10 days (currently 27km to day 8) 33km resolution from 11 to 16 days (currently 55km from day 9 to 16) Physics changes: Radiation modifications -- McICA Hybrid EDMF PBL scheme and TKE dissipative heating Stationary convective gravity wave drag Data Assimilation and Analysis Stochastic physics in ensemble forecast T574L64 EnKF ensembles (27km with 80 members) Observations GPSRO enhancements improve quality control Updates to radiance assimilation Assimilate SSM/IS UPP LAS and MetOp-B IASI radiances New enhanced radiance bias correction scheme Additional satellite wind data hourly GOES, EUMETSAT Upgrades are complex and cover data assimilation, physics, resolution and infrastructure 35
36 GFS Development Test Results 0.008 0.011
GFS Development Test Results Climate Prediction Center 37
38 GFS Development Test Results Weather Prediction Center Extratropical Cyclone Track Errors > 20N and < -20S (June Nov 2014)
GFS Development Test Results Weather Prediction Center 39
40 GFS Development Test Results Hurricane Prediction Center Atlantic Hurricane Track Errors 2012, 2013 and 2014 Seasons
GFS Development Test Results Aviation Weather Center Jet streams prefer GFSP Turbulence guidance prefer GFSP Tropopause height prefer GFSP S. Hemisphere prefer GFSP Convection prefer GFSP by a smidgen Issues with CP over forecast along coastlines, ITCZ, and frontal zones AWC recommends operational implementation AWC thanks EMC and NCO for development, collaboration, and dataflow 41
42 Global Ensemble Forecast System Scheduled Implementation is May 2015 Model: GSM Semi-Lagrangian model Horizontal resolution: Current: T254 (0-192 hours), T190 (192-384 hours) Upgrade: T574 (0-192 hours), T382 (192-384 hours) Vertical resolution: Current: L42 hybrid levels Upgrade: L64 hybrid levels to match GSM and DA Output: Current: every 6-hr for 1*1 degree Upgrade: every 3-hr for 0.5*0.5 degree
43 Commonly Asked Questions: Production Suite Evolution Global systems increase horizontal & vertical resolution GFS satisfies NAM requirements (GFS to 13km @ day 10 Jan 2015) GEFS satisfies SREF requirements (GEFS to 27km @ day 8 Spring 2015) GEFS reforecasts a new requirement (Maintain current GEFS for 1-year) Regional systems shift to convection permitting ensembles HRRRE to satisfy WOF & NAM requirements (GSI, ARW and NMM) Emerging requirements at weeks 3 & 4 and sub-seasonal to seasonal Coupled GEFS extended to 30-days + reforecasts Improve CFS for sub-seasonal to seasonal Operationalize the North American Multi-Model Ensemble (NMME) Coordinate other components of the production suite Hydrologic Space weather Ecological Arctic
Feedback: Model Requirements and Pre-Implementation Assessments Requirements definition Identified as a weakness by NCEP stakeholders incomplete requirements may create false expectations NWS needs an improved process is portfolio management the answer? Stakeholders--- need earlier access to information What changes are being made? What s the rational? What characteristics of the tool will change? Stakeholder calibration methods need time and access to preimplementation data in order to adapt (i.e., GEFS FY15 Upgrade) 30-day NCO parallel insufficient for customer assessment IMPROVE COMMUNICATION BETWEEN MODEL DEVELOPERS AND STAKEHOLDERS 44
What Will This Look Like in 2025? 45
46 Thanks for the interest and opportunity
Backup Machine Used for Development and Transition to Operations 24-h Cycle 6 June 2014 Number of Nodes RAP GFSx GEFS 00 06 12 18 00 Time of Day (UTC) 47
Number of Nodes Why 2 Identical Clusters: Switch Primary to Backup in 15 Minutes 24-h Cycle 8 July 2014 00 06 12 18 00 Time of Day (UTC) 48