Presented by: Vijay Tallapragada Chief, Global Climate and ...
Decisional Brief to NCEP Director GDAS/GFS V13.0.0 Upgrades for Q3FY2016 Presented by: Vijay Tallapragada Chief, Global Climate and Weather Modeling Branch NCEP/EMC March 17, 2016 1 GDAS/GFS upgrade Project Status as of: 3/17/2016 G G Project Information and Highlights
Leads: Vijay Tallapragada, EMC, Becky Cosgrove, NCO Scope: 1) Upgrade to 4D hybrid EnVar data assimilation 2) Produce hourly output out to 120 hrs 3) Address high bias in 2m temp. during summer* Estimated Benefits: 4) Generally more skillful forecasts Estimated Resources: 5) In the process of determining resources Milestone (NCEP) Date Status 6/1/15
Complete Management Briefing 1/15/2016 2/25/2016 3/10/2016 3/17 TODAY Final GFS and all downstream codes submitted to NCO 10/27/15 1/15/2016 1/22/2016 1/27/2016 2/3/16 Complete All non-GFS downstream codes submitted to NCO
2/9/2016 2/19/2016 3/4 Complete Technical Information Notice Issued 11/30/15 2/23/2016 -->4/1 SPA begins prep work for 30 day test 10/28/15 1/19/2016 1/23/2016 1/28/16 2/4/2016 Initial coordination with SPA team Submit frozen codes to NCO to setup real-time and retrospective runs Pre-CCB Briefing to EMC and OD Completion of full retrospective runs EMC testing complete/external evaluation complete
EMC CCB approval Issues/ Risks G Scheduling Issues: 24-hr parallel production test Mitigation: R Management Attention Required Y
2/2/16 4/18 4/29 5/4 5/11 Operational Implementation 2/16/16 4/19 5/3 5/10 5/17 Potential Management Attention Needed G On Target 2 Implementation Overview 1) The GDAS/GFS is being upgraded to 4D-Hybrid En-VAR System
2) 3) 4) 5) Research supported by Sandy Supplemental transitioning to operations Land surface improvements to address summertime warm/dry biases in surface fields Hourly output fields through 120-hr forecasts Evaluation of GDAS/GFS upgrades based on 34 months of retrospective and real-time experiments Revised vertical structure for GFS/GDAS components following NCO Standards 3 4D EnVar: The Way Forward Natural extension to operational Hybrid 3DEnVar
Uses variational approach with already available 4D ensemble perturbations No need to develop or maintain TLM and ADJ models Makes use of 4D ensemble to perform 4D analysis Modular, usable across a wide variety of models (NGGPS dycore replacement) Highly scalable and computationally inexpensive (w.r.t. 4DVar) Aligns with technological/computing advances Estimates of improved efficiency e.g. at Env. Canada (6x faster than 4DVAR on half as many cpus) Combines best aspects of variational and ensemble DA algorithms Other centers exploring similar path forward for deterministic NWP Canada (replaced 4DVAR), UKMO (potentially replace En4DVar) 4
Data Assimilation Upgrades Use 4D covariances instead of 3D Multivariate Ozone update Assimilate all-sky (clear and cloudy) radiances Bias correct aircraft data And other upgrades (e.g. CRTM, Data selection/thinning, AMV winds, etc.) 4DHybrid details Current 3DHybrid Proposed 4DHybrid Static / Ensemble Weights 25% static ; 75% ensemble 12.5% static; 87.5% ensemble Additive Inflation
5% 0% Tropospheric localization length scales of current 3D Hybrid 5 Systematic reduction of RMSE from 4DHYB 6 Multivariate update for Ozone Multivariate => Ozone observations are allowed to update other state variables, and other state variables are allowed to produce ozone increments through cross-covariances O3 RMSE NH
Aircraft Bias Correction Remove warm bias (approx. 200hPa) from aircraft temperature data Also bias correct ascending and descending legs OmF StdDev cruise level Before BC OmF Histograms Before BC After BC ascent After BC descent
OmF Bias 10 Forecast Model and Product Changes Convective gravity wave upgrade Tracer adjustment upgrade Corrections to land surface to reduce summertime warm, dry bias over Great Plains GFS showed too little evaporation and too much sensible heat flux, hence Bowen ratio is too high. Upgraded LSM includes rsmin for grassland from 45 to 20 rsmin for cropland from 45 to 20 roughness length for cropland from 3.5cm to 12.5cm (used to address too strong surface winds) Improved icing probability products and new icing severity product
5 more levels above 10 hPa Hourly output through 120-hr forecast 11 New Model Upgrade Evaluation Strategy Real-Time and Retrospective Parallels: GCWMB real time (pr4devb) period: 2015070100 - real time Evaluation Procedures: GCWMB 2015 summer retrospective (pr4devbs15) period: 2015041500 - 2015120100 (230 days) Involve field in real-time and retrospective evaluation of science upgrades
Identify case studies and provide data for extended evaluation period beyond last 30-day parallel NCO 30-day parallel is only for IT evaluation GCWMB 2013 summer retrospective (pr4devbs13) period: 2013041500 - 2013120100 (230 days) NCO 2013-2014 winter retrospective (pr4devbw13) period: 2013110100 - 2014060100 (212 days) NCO 2014 summer retrospective (pr4devbs14) period: 2014050100 - 2014120100 (214 days)
GCWMB 2014-2015 winter retrospective (pr4devbw14) period: 2014110100 - 2015070100 (242 days) GCWMB Special retrospective for H. Sandy period: 2012101700 - 201213100 (15 days) 12 Comprehensive Scientific Evaluation of GDAS/ GFS Upgrades RetrospectivesStandard verification page against own analyses, GFS2015 vs. GFS2016: http://www.emc.ncep.noaa.gov/gmb/wx24fy/vsdb/gfs2016/ Western Region side by side comparison of operational GFS and parallel GFS: http://ssd.wrh.noaa.gov/gfs/html/ (courtesy: Mark Loeffelbein) Real time plots of near surface variables at representative stations: http://www.emc.ncep.noaa.gov/gc_wmb/parthab/Plume_test/GFSx/EMCGEFSpl umes.html GFS Soundings available on case by case basis, real-time page for selected cities: http://www.emc.ncep.noaa.gov/gc_wmb/tdorian/meg/index.html Several case studies from field; Real-time and retrospective data dissemination
to evaluators Evaluation of downstream models (HWRF, Ensembles, Waves) 13 Case studies and evaluation/analysis by MEG/ EMC, Centers, Regions & Others Case Studies Case Studies MEG review of case studies proposed by WPC, Wester n Region and Central Region MEG review of additional case studies A case study for Dec. 5-6, 2013 requested by Souther n Region Blizzard of January 22-23, 2016 Presentation to WPC on case studies
Precipitation cases for WPC Western Region/Central region case study Height field evaluation for WPC Central Region case study, Alaska case study and Sout hern Region case study Operational and experimental GFS forecasts for Atsan i (extratropical transition, Alaska region) MODE evaluations of new GFS: Precip; Total Winds; Zonal Winds; Meridional Winds; and CAPE Case studies for Central Region March 23, 2015; April 2, 2015; June 4-5, 2015; July 6, 2015 A case study of the Nov. 16-17, 2015 tornado outbreak in Texas and Oklahoma Evaluation from EMC Teams: HWRF; Ensemble; Wave
Hurricane Joaquin and South Carolina flooding WPC documentation of dry bias over the southeast U S in the GFS and GFSX Case study of GFS and GFSX cold bias over snowpac k Verification from Data Assimilation perspective Warm dry bias over Great Plains in summer: Here and Here; Case study: Here MEG presentations reviewing the new GFS Nov. 12 ; Nov. 19; Dec. 17; and Feb. 11 Extratropical storm tracks Evaluation from the Centers: CPC; NHC; SPC; OPC; Comparison of systematic errors in the GFS and GFSX
Forecast tracks for Sandy 14 Evaluation of Q3FY16 GDAS/GFS Upgrade: EMC Perspective 15 Summary of various evaluation metrics Evaluation Remarks Analysis increments 2016 GFS much smaller increments --analysis and first guess in better agreement
Score card Significant improvements in many aspects of the evaluation metrics. Upper Stratospheric biases showed degradation. 500 hPa ACC 0.004 gain in NH; 0.007 gain in SH; statistically significant improvements through 168 hrs Surface heights Significant improvements through 192 hrs in both hemispheres Winds Significant reduction of RMSE through 240 hrs in both hemispheres and global tropics Temperature
RMSE Big improvements in Southern Hemisphere. Upper troposphere/ Stratosphere in Northern Hemisphere has increased RMSE. 850 hPa temperatures significantly improved. Temperature fit to obs Better fit to obs except in the upper stratosphere. Significant reduction of RMSE in NH, SH and global tropics. 16 Summary of various evaluation metrics Evaluation Remarks Vector wind
RMSE Better fit to obs, significant reduction of RMSE in NH, SH at 850 and 200 hPa. No significant change in global tropics. CONUS Precip Rain/no rain (Threshold of 0.2 mm/day) worse in GFSX Thresholds of 2 to 25 mm/day significantly improved CONUS Near Surface Fields Significant improvements in T2m, Td2m, Latent Heat, CAPE and Surface Winds Hurricane Tracks and cyclogenesis
Positive improvements in both NATL and EPAC, for tracks and intensity. Significant improvement in tropical cyclogenesis forecasts. TAFB GFSP seemed to have an advantage at longer lead times for gap wind events Extra tropical cyclone tracks 7 out 10 times, errors in GFSX are smaller than in GFSO in winter. During summer months, the errors are always smaller in parallel GFS. OPC Evaluation Track errors for winter season are a slight improvement shorter
term and no significant improvement medium range 17 Summary of various evaluation metrics Evaluation Remarks MODE verification Jet Streams: GFSX generally looks better and closer to the ECMWF; QPF: GFSX has higher MMI (Median of Maximum Interest) values for all forecast hours except at 60-h; CAPE: GFSX somewhat better than GFS. Both underestimate compared to RAP analysis Case studies from Field
GFSX better in 6 cases out of 9, operational GFS better in 3 (subjective evaluation) Typhoon Astani GFSX better in 7 verification times, operational GFS in 3 verification times. WPC Case studies Of the 6 precipitation case studies (36 hour forecasts), the GFSX did better for 3 cases, the operational GFS was better for 1 case, and both models tied for 2 cases. Ensemble Team verification
2014 Winter: Good for short forecast (days 1-3); Slight degradation (days 5-10). 2014 Summer: Good for all lead time (out to day 12) HWRF Team New GFS shows improved track and intensity forecasts in the N. Atlantic and neutral impact in the E. Pacific 18 2016 GFS much smaller increments -analysis and first guess in better agreement DA Impact Highlights: Better Fit-to-obs for Temperature and Winds at various levels
Improved minimization Significant improvement in the shortrange forecasts for several variables 19 Improved Fit to Obs 20 Anomaly Correlations & RMSE: GFSX vs. GFS; 34 months verified against own analyses 21 Biases: GFSX vs. GFS Significant improvements in many aspects of the evaluation metrics.
25 RMSE for Winds: Northern Hemisphere 200 hPa 850 hPa 26 RMSE for Winds: Southern Hemisphere 200 hPa 850 hPa 27 RMSE for Winds: Global Tropics
200 hPa 850 hPa 28 RMSE for Temperature 29 CONUS Precip ETS (00Z & 12Z) 30 CONUS Precip Skill Scores 31 RMS errors for selected atmospheric fields verified against own analyses reduced
Forecast length Day 1 Day 3 Day 5 Day 8 Extra-tropics 10% 4% 2% 1% Tropics 10% 4% 2% Period of verification: May 2013-Feb. 2016) Over US 96 hr RMS errors of near-surface fields against station observations reduced (4 times/day, complete4 annual cycle
2 m Temperature 2 m dew point 10 m wind WEST 2% 2% 4% EAST 3% 3% 8% Assessment of impact of LSM changes 2m T cooler, significant improvements over southern plains and southeast. Bias is worse over the Northern Plains and Northeast. RMS errors significantly improved over northern and
southern plains, Southeast and Alaska, worse over northwest 10 m winds decreased, RMS error improved The land surface parameter refinements have significantly reduced the warm/dry biases in the summer The change has little impact in the winter. However there are some degradations in the spring/fall. Also it is worst in 00Z (sunset). Some of them will be addressed in the next GFS physics implementation. 33 Significant improvement of biases for near surface variables 34 Case Studies from the Field: EMC Evaluation Case CR 1/29-2/2/2015 WR 10/3-10/4/2015 WR 11/8-11/10/2014
WR 11/20-23/2014 WR 8/28-8/30/2015 AK Typhoon Astani Case SR 12/5-12/6/2013 CR 3/23/2015 CR 6/4-6/5/2015 CR 7/6/2015 GFSX 7/10 Model Performance GFSX somewhat better GFSX slightly better GFS slightly better GFSX better GFSX slightly better GFSX better Model Performance
GFSX did better GFSX did slightly better GFS did slightly better GFS did slightly better GFS 3/10 35 Impact on Hurricanes: NHC evaluation of Track and Intensity Errors 36 Impact on Hurricanes: NHC evaluation of Track and Intensity Skill 37 Impact on Hurricanes: Frequency of Superior
Performance for Track Forecasts 38 Impact on Hurricanes: Frequency of Superior Performance for Intensity Forecasts 2012-2016 Frequency of Superior Performance - Intensity Atlantic East Pacific 39 Impact on Hurricanes: Evaluation of Tropical Cyclogenesis Forecasts Verification of TC cyclogenesis in the GFSX comparison to current and previous version of the GFS (courtesy of Dan Halperin and Bob Hart)
40 Impact on Hurricanes: NHC Evaluation AL Track Intensity 0-48 h - 3% +5% 72-120 h +7%
+ 11% EP Track Intensity 0-48 h +5% +5% 72-120 h +1% +2%
Track and intensity error improvements/degradation of Q3FY16 GFS vs. 2015 GFS for the 2012-2016 retrospective runs, by basin NHC Evaluation Report: GFSP has mostly improved TC track, intensity and genesis forecasts in comparison to current GFS TAFB Evaluation Report: GFSP in general handles gap wind events a little better than the current GFS, especially at longer time ranges. 41 Impact on Extratropical Cyclone Tracks: Winter Guang Ping Lou 42 Impact on Extratropical Cyclone Tracks: Summer Guang Ping Lou
43 WPC Evaluation of Case Studies WPC Case Studies Remarks Tornado outbreak over Kansas, Texas Nov. 16-17, 2015 GFSX better in forecast from 000 GMT Nov. 16 Sandy Oct .22-30, 2012 GFS, GFSX track errors similar Joaquin Sept. 25-Oct. 4, 2015
GFSX better track, adopted out to sea track 6 hours before operational GFS South Carolina flooding Oct. 3, 4, 2015 GFS, GFSX similar GFS dry bias in southeast US autumn 2015, winter 2015-2016 GFS, GFSX similar GFS cold bias over snow cover GFS, GFSX similar Blizzard Jan. 22-23, 2016 GFS, GFSX similar
Warm, dry bias Great Plains 000 GMT Aug. 16 GFSX better New England blizzard Jan 26-27 2015 GFSX better 2.5 day forecast 44 Impact on Operational HWRF Track and Intensity in the Atlantic and EPAC basins 45 Endorsements from Stakeholders Region/Center Recommendation Remarks
Western Region Implement Neutral Central Region Implement with reservations Little improvement Southern Region Implement No striking differences Eastern Region Implement
Minor improvements Pacific Region Implement Models performed well with Winston Alaska Region Implement No specific problems WPC Implement
Similar, GFSX slightly better sometimes NHC Neither endorse nor oppose Improved tropical forecasts, downstream tests for HWRF incomplete (70% of the retrospectives completed as of today) 46 Endorsements from Stakeholders Region/Center Recommendation Remarks
Implement Need improvements in upper atmosphere MDL Implement Redeveloped MOS better NWC Implement Hourly files should improve NWC fcsts SPC
Implement Improved in warm season Weather It Is Ltd. under situations where the observational network is more (Prof. Barry Lynn) dense, there has been improvement in the initial state (and lateral boundary conditions) of the GFSX compared to GFS AccuWeather Hourly output is of significant value for Weather Industry 47 Hourly Output from GFS through 120 hrs & Additional Fields Hourly GFS forecast output at 0.25 deg. resolution (grib2) will be made available through 120 hr (ftp only) GFS Post is adding output on 5 more pressure levels in
stratosphere 1, 2, 3, 5, and 7 mb per request of CPC. Variables include Geopotential Height (HGT); Temperature (TMP); Relative Humidity (RH); U- and V Components of Wind (UGRD & VGRD); and Ozone Mixing Ratio (O3MR) Two New Products: Icing probability and Icing Severity are also added to Aviation Weather (WAFS) 48 Feedback from AccuWeather Increasing the temporal resolution of the model output will help provide more detailed timing forecasts of precipitation, enhance the way in which extreme weather events are forecast, and provide this granularity on a global basis, key to an interconnected world. Hourly output will provide an additive piece of information to use in formulating forecasts, providing end users with higher resolution details at when precipitation may begin or end. Hourly data can be used by logistics operators, transportation companies, business travelers, general users, commodity companies, agricultural interests, and the general public to better plan their day and make informed decisions.
We applaud the effort by NCEP to release this hourly data and are excited to integrate it into our forecasts. The increase in temporal and spatial resolution of model guidance is key to enhancing weather forecasts and their contextual relevance to users. 49 EMC/GCWMB Assessment Positive evaluation (significantly positive improvements in majority of the metrics) DA upgrades have been effective in reducing the forecast errors in the short-range, and improving analysis increment for almost all prognostic variables Results showed significant improvement in week 1 forecasts verified against own analyses except for heights and temperatures in stratosphere Rain no rain forecasts worse, but overall conus precipitation improved significantly 50
EMC/GCWMB Assessment 2m temperature, dewpoint, 10 m wind forecasts against station obs over CONUS, Alaska improved. CAPE forecasts over CONUS improved Forecasts of tropical storm genesis. track and intensity forecasts improved. Mode verification of CAPE, Jet Streams, QPF and winds shows GFSX slightly better Synoptic evaluations of GFSX produced no red flags. GFSX, GFS similar; GFSX slightly better in some cases Forecasts of heights, temperatures, winds significantly improved except for heights and temperatures in stratosphere. Large errors in upper stratosphere 51 Resource Changes for GFS/GDAS 52
Revised Vertical Structure for GFS/GDAS 53 Special acknowledgements John Derber, Russ Treadon, Glenn White, Fanglin Yang, Tracey Dorian, Partha Bhattacharjee, Lin Gan, Boi Vuong, Qingfu Liu, Guangping Liu, Diane Stokes, Dennis Keyser, Yali Mao, Eugene Mirvis, George Gayno, Zhan Zhang, Lin Zhu, Cathy Thomas, Ed Safford, Rahul Mahajan, Jeff Whitaker, Yuejian Zhu, Hendrik Tolman, Mike Ek, Helin Wei, Jesse Meng, Steven Earle, Jen Yang & Becky Cosgrove MEG Team, and all evaluators from various centers and regions 54
EMC/GCWMB requests NCEP Director to approve implementation of Q3FY16 GDAS/GFS upgrade package. 55 Next Steps Code Hand-off to NCO: Completed All non-GFS downstream codes submitted to NCO: Completed Collect Evaluation Reports from the field: Completed Final EMC CCB: Today (Completed) OD Briefing: 3/17/16 (Completed) TIN: 4/1/2016 (on track) 30-day evaluation: 4/06 5/5 Final OD Briefing by NCO: 5/11 Implementation: 5/17/16 56
Lessons Learned Availability of computational (and human) resources posed significant challenges (disk, archive, data transmission, graphics and visualization capabilities etc.) New implementation strategy helped extending the scientific evaluation beyond 30-day NCO parallels Project Management approach helped us (EMC and NCO) keep the schedules once we reached critical decision making points EMC Model Evaluation Group (MEG) worked diligently with all the evaluators for timely feedback and responses Evaluation extended beyond traditional metrics, with emphasis on sensible weather elements and specific case studies, involving over three years of retrospective and real-time data. Expectations from the field far exceeded the scope of model upgrades targeted for this implementation 57
Laboratory, Workshop, Drawing Office, CAD and other CEM. Project work. Individual. Group e.g. EGPR, Formula Student, Dev Tech, UAV Design. What about Placement Opportunities? Available after Part 2 or Part 3. Placements in the last two years include. Airbus. British...
Source: WVU. If workers must access the top of a tank to read a gauge. Moving the tank gauge to ground level would eliminate the need for fall hazard exposure. As a worker, you could suggest such. Using a drone...
1. Find the degree of a polynomial. Arrange the terms of a polynomial in ascending or descending order. What does each prefix mean? mono one bi two tri three What about poly? one or more A polynomial is a monomial...
2. Safety: Feeling physical safety and security, stability, familiar surroundings, and so forth are manifestations of safety needs. They are aroused after physiological motives are minimally satisfied, and before other motives. 1. Physiological: Food, water, sleep, and to a limited...
EDMS Gradebook: a PADI Application Larry Hamel CodeGuild, Inc. AERA April 2005 EDMS Gradebook Purpose Gradebook combines assessment specification with student scores, and employs a scoring engine to finally display estimates of student proficiency Gradebook Purpose (cont.) Gradebook keeps track...
By Hannah, Allyssa, Alondra (Nick didn't do anything) Description. Food photography is a still life photography genre used to create attractive still life photographs of food. It is a specialization of commercial photography, the products of which are used in...
Loaded Word Use "loaded" words like . . . new improved best Misuse of Statistics When the statistics are based on a falsehood. Name Calling or stereotyping Snob Appeal aims to flatter Makes insinuation that this product is better than...
Ready to download the document? Go ahead and hit continue!