US THORPE 19 Sept 2012 1 N C E P Ensemble Systems within the NOAA Operational Modeling Suite William. M. Lapenta Acting Director Environmental Modeling Center NOAA/NWS/NCEP
Data Assimilation, Modeling and High Performance Computing Necessary to Meet NOAA/NWS Operational Requirements Boxes marked with a indicate where a Data Assimilation and Modeling numerical guidance system directly supports a NWS GRPA Goal NWS Government Performance and Results Act Goals NWS GPRA Goals Flash Flood Warnings Lead Time Flash Flood Warnings Accuracy Marine Wind Speed Forecast Accuracy Marine Wave Height Forecast Accuracy Aviation Forecast IFR Accuracy Aviation Forecast IFR False Alarm Ratio Winter Storm Warnings Lead Time Winter Storm Warnings Accuracy IMET Fire Response Time Precip Forecast Day 1 Threat Score US Seasonal Temp Forecast Skill Hurricane Forecasts Track - 48 hr Error Hurricane Forecasts Intensity - 48 hr Error Tsunami Message Response Time Geomagnetic Storm Forecast Accuracy Geomagnetic Storm Forecast False Alarm Ratio NCEP Production Suite Numerical Guidance Systems HYSPLIT RAP NAM SREF GFS GEFS NAEFS CFS HYCOM WW3 Wave Tsunami HWRF GFDL ENLIL Ensemble US THORPE 19 Sept 2012 2
Forecast Lead Time US THORPE 19 Sept 2012 3 Seamless Suite of Numerical Guidance Systems Spanning Weather and Climate Outlook Seasons Years Forecast Uncertainty Guidance Threats Assessments Forecasts Watches Warnings & Alert Coordination Minutes Hours Days 1 Week 2 Week Months Climate Forecast System North American Ensemble Forecast System Global Ensemble Forecast System Global Forecast System Short-Range Ensemble Forecast North American Mesoscale Hurricane WRF & GFDL Rapid Refresh for Aviation Dispersion Models for DHS Benefits Waves Ocean Forecast System Space Weather Tsunami
Number of Nodes 0:00:00 0:45:00 1:30:00 2:15:00 3:00:00 3:45:00 4:30:00 5:15:00 6:00:00 6:45:00 7:30:00 8:15:00 9:00:00 9:45:00 10:30:00 11:15:00 12:00:00 12:45:00 13:30:00 14:15:00 15:00:00 15:45:00 16:30:00 17:15:00 18:00:00 18:45:00 19:30:00 20:15:00 21:00:00 21:45:00 22:30:00 23:15:00 US THORPE 19 Sept 2012 4 Production Suite on Supercomputer STRATUS HWM January 2010 Node use for 20091202 140 130 Development Work dev ofs_atl test 120 para 110 100 multi misc cdc 90 80 Fence dgex aqm cdas 70 60 hiresw High Water sref Mark godas 50 cfs 40 30 mrf firewx merge 20 ruc2 10 hur wave 0 00 06 12 18 00 gefs gdas gfs_ana ysis gfs Time of the day (utc) ndas nam
Number of Nodes US THORPE 19 Sept 2012 5 Production Suite on Supercomputer August 2012 High Water Mark 2010 00 06 12 18 00 Time of the day (utc)
NOAA Weather and Climate Operational Supercomputer System (WCOSS) IBM idataplex Systems Bridge System Life Cycle Arch OS Average Capability Oct 2012 - Sep 2013 Average Capacity Nbr of compute / batch cores TeraFLOP Power6 AI 5.3 1.0 1.0 5,314 73.9 TF WCOSS Phase 1 Plan accept Dec 2012 FOC Aug 2013 idataplex Linux (RHEL) 2.0 over Bridge P6 2.3 over Bridge P6 10,048 208 TF Located in Reston VA and Orlando FL Schedule: Full system available in September 2012 Acceptance of both systems Jan 2013 Operational no later than 1 October 2013 US THORPE 19 Sept 2012 6
Dual-Resolution Coupled Hybrid 3D-VAR/EnKF member 1 forecast T254L64 member 2 forecast Uses background error covariances computed from the ensemble EnKF member update Generating new ensemble perturbations given the latest set of observations and a first-guess ensemble EnKF ensemble perturbations are "re-centered" around the highres analysis member 1 analysis member 2 analysis member 80 forecast T574L64 forecast first-guess ensemble used to estimate background error covariances GSI Hybrid Ens/Var Replaces the EnKF ensemble mean analysis analysis Deterministic forecast member 80 analysis Used for GFS forecasts for next cycle Previous Cycle Current Update Cycle US THORPE 19 Sept 2012 7
US THORPE 19 Sept 2012 8 Global Data Assimilation System Upgrade Hybrid system Most of the impact comes from this change Uses ensemble forecasts to help define background error NPP (ATMS) assimilated Quick use of data after launch Use of GPSRO Bending Angle rather than refractivity Allows use of more data (especially higher in atmos.) Small positive impacts Implemented 22 May 2012 Satellite radiance monitoring code Allows quicker awareness of problems (run every cycle) Monitoring software can automatically detect many problems Post changes Additional fields requested by forecasters (80m variables) Partnership between research and operations
Global Model Track Guidance for TS/Hurricane Isaac US THORPE 19 Sept 2012 Sequence of 5-day 00 and 12 UTC cycle Isaac forecasts 26 August 2012 to 29 August 2012 NCEP GFS ECMWF
Global Ensemble Systems Parameter NCEP CMC NAEFS Model GFS GEM NCEP+CMC Initial uncertainty ETR EnKF ETR + EnKF Model uncertainty/stochastic Yes (Stochastic Pert) Yes (multi-physics) Yes Tropical storm Relocation None Daily frequency 00,06,12 and 18UTC 00 and 12UTC 00 and 12UTC Resolution T254L42 (d0-d8)~55km (d0-d16) ~ 66km 1*1 degree T190L42 (d8-16)~70km Control Yes Yes Yes (2) Ensemble members 20 for each cycle 20 for each cycle 40 for each cycle Forecast length 16 days (384 hours) 16 days (384 hours) 16 days Post-process Bias correction (same bias for all members) Bias correction for each member Yes Last Upgrade February 14 th 2012 August 17th 2011 GEFS Paper: 4B3.4 Thursday at 11:15: Yuejian Zhu et al. (Fanglin Yang presenting) US THORPE 19 Sept 2012 10
US THORPE 19 Sept 2012 North American Ensemble Forecast System 8 to 14 Day Temperature Guidance Experimental Valid: 27 January to 02 February 2012 Issued: 19 January 2012
Advancement of Climate Forecast System (CFS) Implemented 30 March 2011 Attribute CFSv1 (Operational 2004) CFSv2 (March 2011) Analysis Resolution 200 km 27 km Atmosphere model Ocean model Land surface model (LSM) and assimilation 200 km/28 levels Humidity based clouds MOM-3: 60N-65S 1/3 x 1 deg. Assim depth 750 m 2-level LSM No separate land data assim 100 km/64 levels Variable CO2 AER SW & LW radiation Prognostic clouds & liquid water Retuned mountain blocking Convective gravity wave drag MOM-4 fully global ¼ x ½ deg. Assim depth 4737 m 4 level Noah model GLDAS driven by obs precip Sea ice Climatology Daily analysis and Prognostic sea ice Coupling Daily 30 minutes Data assimilation Retrieved soundings, 1995 analysis, uncoupled background Radiances assimilated, 2008 GSI, coupled background Reforecasts 15/month seasonal output 25/month (seasonal) 124/month (week 3-6) US THORPE 19 Sept 2012 12
National Multi-Model Ensemble (NMME) Project Facilitated by the NOAA Climate Test Bed NMME as a Modeling Test-Bed Seasonal to Interannual Time Scales Predictability Research: e.g., South East US Drought Model Evaluation and Development Initialization Strategies: e.g., Land, Ocean Fosters interaction between research and operations Provides experimental guidance products to Climate Prediction Center Participating Organizations: University of Miami - RSMAS National Center for Atmospheric Research (NCAR) Center for Ocean-- Land-- Atmosphere Studies (COLA) International Research Ins7tute for Climate and Society (IRI) Canadian Meteorological Centre (Soon) NASA GMAO NOAA/NCEP/EMC/CPC NOAA/GFDL Princeton University University of Colorado (CIRES Data are available at: http://iridl.ldeo.columbia.edu/sources/.models/.nmme/ US THORPE 19 Sept 2012 13
Current Implementation Strategy for Operational Global Systems 2009-20012 development and upgrade cycles for the GFS, GDAS and GEFS System GFS GDAS GEFS 2009 2010 2011 2012 J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J HURRICANE HURRICANE HURRICANE HURRICANE Computer T574 Adj T574 LSM Fix T382 3dvar+Obs 3D-EnVar T190L28 T254L42 Computer HURRICANE HURRICANE HURRICANE HURRICANE J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J 2009 2010 2011 2012 GFS, GDAS and GEFS have separate development and upgrade cycles Advantages: Focused development cycles reduced dependencies Higher probability of meeting planned schedules Stakeholder evaluation focused on single system Computational footprint of reasonable size Disadvantages GEFS uses t-1 or t-2 versions of GFS Multiple 30-day NCO/Customer evaluation periods required (best case scenario may be longer) Redundant 80-member ensembles for ETR and EnKF Implementation Month 30-day NCO Parallel EMC Hard Parallel Development US THORPE 19 Sept 2012 14
Consolidate Development and Upgrade Cycles for the GFS, GDAS and GEFS? System GFS GDAS GEFS Compare 2011-2012 and proposed 2014 upgrade cycles 2011 2012 2013 2014 J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D HURRICANE HURRICANE HURRICANE HURRICANE Computer LSM Fix T878L64 3D-EnVar T574Ens+obs T254L42 T382L64 Computer HURRICANE HURRICANE HURRICANE HURRICANE J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D 2011 2012 2013 2014 Combine GFS, GDAS and GEFS into a single development and upgrade cycle Advantages: GEFS uses the operational version of the GFS Use EnKF to support GEFS and Analysis (remove 80-member ETR ensemble) Forces communication and coordination across global, ensembles and analysis teams plus stakeholders and downstream products Single 30-day NCO/Customer evaluation periods required Disadvantages Stakeholder evaluation focused on multiple systems Computational footprint becomes much larger may need to consider parallel suite testing burden on NCO and development Increased reward/risk performance in one component impacts others Implementation Month 30-day NCO Parallel EMC Hard Parallel Development US THORPE 19 Sept 2012 15