1

Recent Advances at the National Centers for Environmental

Prediction

“Where America’s Climate and Weather Services Begin”

Louis W. UccelliniDirector, NCEP

MIC-HIC WorkshopMay 12, 2004

2

Overview

• NCEP Organizational Status• Status of Models• Recent Advancements• Future Plans for Community Models• Status of Support for Special Projects

– IFPS• DGEX• HPC Products

– Winter Weather Experiment

3

NCEP Organizational Status

4*As of 10/1/04*54 FTE

Total FTE: 429*131 Contractors/24 Visitors

NCEP’s Future is Built Upon:

Climate-Weather-Water-Land-Chemistry Linkages; for example Seasonal Hurricane Outlooks & Extratropical Storm patterns Meteorological-Hydrological forecasts Ocean and atmosphere coupled forecasts Atmosphere-Land Processes coupled forecasts Relationship of solar activity on service provision and climate fluctuations Ozone forecasts by combining air chemistry and operational models

“Seamless Suite” of products through a collaborative approach Extension of predictability of Weather and Climate (from

snowstorms to ENSO); Improve the forecasts of Extreme Events Community Model Approach – Common Model Infrastructure Addressing uncertainty in forecasts – Ensemble modeling

NEW Collaborative Forecasting Unified Model Infrastructure applied from the “Sun to the Sea”

6

7

Status of Models

8

Computing Capability

Commissioned/Operational IBM Supercomputer in Gaithersburg, MD (June 6, 2003)

$20M/Year $20M/Year InvestmentInvestment

•Receives Over 116 Million Global Observations Daily•Sustained Computational Speed: 450 Billion Calculations/Sec•Generates More Than 5.7 Million Model Fields Each Day•Global Models (Weather, Ocean, Climate)•Regional Models (Aviation, Severe Weather, Fire Weather)•Hazards Models (Hurricane, Volcanic Ash, Dispersion)•Upgraded 2.5x by October 1, 2004•Backup for operational side: Fairmont, WV installation Fall’04

9

NCEP Operational ModelsEta

12 km, 60 levels, 84 hrs at 0 , 6, 12 and 18Z

Global Forecast System (GFS)

T254 (~55 km) to 3.5 days (84 hrs), 64 levels

T170 (~75 km) to 7.5 days (180 hrs), 42 levels

T126 (~105 km) to 16 days (384 hrs), 28 levels

16 days (384 hrs)/4 times per day

RUC

20 km, 50 levels

12 hrs at 0,3,6,9,12,15,18,21Z

3 hrs at 1,2,4,5,7,8,10,11,13,14, 16,17,19,20,22,23Z

Climate

T62 (~200 km), 28 levels, 7 months (20 members)

Ensembles

global 10 members at 00, 06,12,18Z

T126 (~105 km) to 180 hrs, T62 (210 km) to 384 hrs

28 levels, 16 days (384 hrs)

regional 10 members at 0 and 12Z

48 km, 45 levels, 63 hrs from 9 and 21Z

Wave Model

global - 1.25 x 1.0 deg lat/lon

Alaskan Regional - .5 x .25 deg lat/lon

Western North Atlantic - .25 x .25 deg lat/lon

Eastern North Pacific - .25 x .25 deg lat/lon

1 level, 168 hrs/4 times per day

North Atlantic Hurricane (seasonal)

North Pacific Hurricane (seasonal)

.25 x .25 deg lat/lon

1 level

78 hours/4 times per day

GFDL Hurricane Model

coupled ocean-atmosphere

Two nests (0.5, 1/6 deg lat/lon)

42 levels

126 hrs at 00, 06, 12 and 18Z

10

High Resolution Applications

• Six Non-hydrostatic Mesoscale

Model (NMM) High resolution Window nested runs (all 8 km except 10 km Alaska) - once per day to 48 hours

• Downscaled GFS with Eta Extension (DGEX) – 12 km, 60 lvl (downscaled

locally to 5km using SmartInit scripts)

– CONUS at 06 and 18Z, AK at 00 and 12Z

– 84-192 h, 6-h frequency

11

High Resolution Applications

• Fire weather runs for IMET

support – 8 km NMM runs in one of 26 areas of coverage, each about 900 km square up to 4/day; area selected by Boise National Interagency Fire Center -SPC

• Dispersion model run on demand using 4 km NMM for Homeland Security

12

Fire Weather IMET Support

8 km NMM captures CA coastal windsGreen – model windsRed – observed winds

12 km Eta 8 km NMM

13

Recent Advancements

• General Performance• Hurricanes• Wave Watch III• QPF• Climate Model

14

15

TPC Atlantic 72 hr Track Forecast Errors

16

NATIONAL HURRICANE CENTER ATLANTIC TRACK FORECAST ERRORS

NATIONAL HURRICANE CENTER ATLANTIC TRACK FORECAST ERRORS

12 24 36 48 72 96 120

Forecast Period (hours)

0

100

200

300

400

500

Err

or

(nau

tica

l mile

s)

1964-1973

1984-1993

1974-1983

1994-2002

Isabel

2003

17

• New model design with emphasis on transparency, vectorization and parallelization (plug compatible, portable).

• More general governing transport equation, allowing for later full coupling with ocean models.

WAVEWATCH III

new model required

• All models use GFS and ice edge information from NCEP's operational ice analysis. A special GFDL driven version of the Western North Atlantic and Eastern North Pacific wave model are run for hurricane wave prediction (72h forecast).

18

Isabel 18/9/2003, 12 UTCnowcast

48h forecast24h forecast

12h forecast

Intensity and location of forecast waves consistent and confirmed by altimeter and buoy observations. At 48h forecast lower wave heights due to earlier landfall.

wave height 50+ ft (45+ ft)

Isabel at Field Research Facility Duck NC

pictures from US Army Corps Of Engineers Field Research Facility webcam

9/18 14:00 EDT 9/29 14:00 EDT

Maximum observed wave height at the end of the pier 26.6ft, which is roughly the maximum sustainable wave height for the local water depth. Wave height 2 miles offshore reported up to 49 ft.

20

Eta Model Performance MetricRatio of Annual 48h Precipitation Threat Score to 24h Score in 1999

21

22

Recent Model Changes

• Regional - Upgraded Eta Physics and Data Assimilation

• Global - Improved Global Forecast System Surface Drag to Improve Synoptic Weather Forecasts

• Ensembles – Increased Global Ensemble Forecast System runs from 2 to 4 per day (10 members each); Extended T126 (~105 km) Resolution from 84 to 180 Hours

• Ensembles - Implemented Bias-Corrected QPF Ensemble Product to add skill to QPF guidance in the 3-5 day range.

• Ocean - Implemented Assimilation of Wave Height Data into Global Wavewatch III Model; Extended runs to 180 hours from 168

• Satellite Data Assimilation - AIRS Assessment leads to Operational Use of AIRS Data by 1st Quarter FY05

23

Recent Advancements: Climate Model

24

Climate Model

• Current operational climate model– 200 km, 28 levels, runs to 7 months each month– Linked to SSTs in Pacific basin only

• Improved operational climate model– Fully coupled ocean-atmosphere system

• NCEP operational Global Forecast System (GFS) atmospheric model

– 200 km resolution, 64 levels, model top 0.2 mb

• MOM3 ocean model (GFDL)– 100 km resolution, 40 levels, 30 km between 10 deg N and 10 deg S– Global; between 65 deg N and 75 deg S– Global Ocean Data Assimilation System (GODAS)

25

Coupled Model Simulation ENSO SST cycles

Nino 3.4 SSTAnomalies

Simulated 2002-2040 (top)

Observed 1965-2003(bottom)

26

Coupled Model Simulation SST Interannual Variability

Observed

64 Level Atm

28 Level Atm

27

Examples of ENSO eventsSimulated El Nino 2015-2016 Simulated La Nina 2017-18

Real El Nino 1982-1983 Real La Nina 1988-1989

28

Ensemble Mean

CMP14CASST

April ICEnsemble Mean – mean of 5 member ensemble

CMP14 – operational dynamical forecast

CASST – Constructed Analog SST (statistical forecast used by CPC)

29

January IC

CASST

CMP14

Ensemble Mean

Ensemble Mean – mean of 5 member ensemble

CMP14 – operational dynamical forecast

CASST – Constructed Analog SST (statistical forecast used by CPC)

30

Recent CFS Runs

April 1 run shows strongwarming event as compared toMar 21 run

31

Most Recent CFS

Latest run shows a more moderate warming

Latest CPC Forecast: ENSO-neutral conditions are expected to continue during the next three months. There is considerable uncertainty about what will happen after July 2004.

32

Future Plans for Community Models

33

Future Plans• WRF

– Common model infrastructure for mesoscale (NCAR and NCEP Dynamic Core and Physics)

– Sustained by AF, Navy, NCEP, NCAR– Testing underway of all combinations

of 2dynamic cores and 2 physics packages at DoD Major Shared Resource Center (one month from each season)

– First operational implementation at NCEP by Oct ’04, implementation at AF by Spring, ‘05

•ESMF–Global common model infrastructure–NCAR, GFDL, NASA/GSFC, MIT, NCEP–Basis for next generation global data assimilation and forecast system–Superstructure and infrastructure nearly complete

NMM

NCAR/MASS

NavyCOAMPS

34

Near - Future Plans• Ensemble models

– SREF (10 members twice/day, 48 km, 45 levels, 63 hrs)– Global (10 members 4/day; 105 km to 180 hrs, 210 km to

384 hrs; 28 levels)

GOAL: To create a North American Ensemble Forecast System with the Canadian Meteorological Centre

Dominant Precip Type63 hour forecast

Valid 12Z,December 5, 2002

35

Support for Special Projects

• IFPS– DGEX– HPC Products

• Winter Weather Experiment

36

Downscaled GFS with Eta Extension (DGEX)

• 12 km, 60 level (downscaled locally to 5km using SmartInit scripts)

• CONUS at 06 and 18Z, AK at 00 and 12Z• 84-192 h, 6-h output frequency• New DGEX grids will be developed and added to production

with content to be defined by the ISST.

• SBN’s TG2 channel to carry DGEX data using GRIB2– AWIPS OB3.2 will have GRIB2 decoding capability

• Hawaii and Puerto Rico domains to be added

37

Expanded HPC Medium-Range Support

• HPC is working to implement the NWS Corporate Board Decision• HPC forecasts at 390 MOS points are available in text format• Also available are HPC forecasts on a 5-km grid using information from ~6000

MOS points and PRISM to add detail to temperatures• Products available now on HPC ftp site:

– max/min temperatures – 12-hour PoP

• Estimated dates for additional fields– dewpoint - Jun 2004– sky cover - Aug 2004– wind direction & speed - Sep 2004

• Problematic fields: weather type, intensity, & character• WFOs can develop smart tools to incorporate local effects• WFOs modify HPC grids at their discretion• WFOs populate the NDFD

HPC day 5 minimum temperature forecast

valid May 4, 2004

38

Winter Weather Experiment 3 (2003-2004)Overview of Participants & Activities

•Two WFO sectors

• Intermountain (IM) - GREEN

• Non IM - RED

• Total of 77 WFOs

•HPC produced WWE graphics

•Available via HPC website

•WFO-HPC Collaboration

•Scheduled ~15-minute call twice daily

•Also via 12 Planet

39

Winter Weather Experiment 3 (2003-2004) WFO feedback implications for next winter

• HPC internal products for WFOs should include– Visualization of model preferences, trends, and winter-

weather-related diagnostics NOT already available at WFOs

– Accumulation graphics of S/IP & ZR like previous WWEs but in 24h increments out to 72h

– Notification of product availability via 12 Planet• WFO-HPC collaboration should occur primarily in 12 Planet

– WFOs prefer to collaborate initially WFO to WFO using HPC products as a reference

– HPC-WFO calls should be reserved for when collaboration via 12 Planet is not efficient

• Initiated by WFO– HPC testing 5-minute recorded “streaming”

video map discussion for internal NWS use• HPC public product issuance deadlines

should be changed – To allow more time to prepare internal

NWS products and collaborate with WFOs

Probability of 4-7"

Probability of 8-11" Probability of 12+"

Expected Amounts

24

Streaming Video Test

A Proposed 4-panel Public Product

401998

41

Summary• NCEP is positioned to deal with important

strategic issues– Climate-weather-water linkage– Expand into “environmental” Air Quality prediction

(with EPA)– Extend predictive capabilities through week 2– Extend consistent predictive capabilities for extreme

events out to Day 7 – Support WFOs/RFCs with an expanded array of

gridded products

42

Summary (cont)• Based on Partnership with larger research community

– Community model approach (global and regional)– Active participation in field programs

• North American Monsoon Experiment• THORPEX

– Test Beds:• USWRP/Joint Hurricane Test Bed (TPC)• Hazardous Weather Forecast Test Bed (SPC) • Aviation Test Bed (AWC)• Proposed USWRP/Hydrometeorological Test Bed (HPC)

– Data Assimilation efforts through JCSDA (Appendix)

43

Appendix

44

45

N.H. 500 mb Height Anomaly Correlation for Forecasts Days 3 (blue), 5 (aqua), and 7 (red)

Monthly Values and Annual Averages

46

S.H. 500 mb Height Anomaly Correlation for Forecasts Days 3 (blue), 5 (aqua), and 7 (red)

Monthly Values and Annual Averages

Status of Distributed ModelsThe Workstation Eta

A means for providing real-time high-resolution numerical model data at the local level

Domain can be placed anywhere on the globe: size and resolution determined by user

Non-NWS use encouraged. About 140 international requests from countries such as China and Brazil (both with >5 users), Turkey and Thailand.

Over 155 domestic users: WFOs, researchers and students at U.S universities

http://www.emc.ncep.noaa.gov/mmb/wrkstn_eta/

48

Hindcast Skill Assessment

• Methodology– 5-member ensemble over 22 years from 1981-2002

– January and April initial conditions

– 9 month runs– Initial atmospheric states 0000 GMT 19, 20, 21, 22, and

23 for each month– Initial ocean states NCEP GODAS (Global Ocean Data

Assimilation System) 0000 GMT 21st of each month• Forced by Reanalysis 2 parameters

• Preliminary results– 10-15 member ensembles for full calibration runs

(ongoing)

49

Hindcast Skill Assessment (cont)

• Hindcast skill– Estimated after bias correction for each year– Uses model climatology based on the other years– Anomaly correlation skill score for Nino 3.4 region

SST prediction– Skill maps for

• Global SST• U.S. temperature • U.S. precipitation

– Comparisons with • Operational dynamical forecast (CMP14)• Operational statistical forecast (NCEP CPC tools)

– Constructed Analog SST (CASST)

50

JCSDA

Count

(Mill

ions)

Daily Upper Air Observation Count

2002

2003

51

Joint Center for Satellite Data Assimilationcreated July 2, 2001

Increase uses of current satellite data in NWP models Develop the hardware/software systems needed to assimilate data from

the advanced satellite sensors Advance the common NWP models and data assimilation infrastructure Develop common fast radiative transfer system Assess the impacts of data from advanced satellite sensors on weather

and climate predictions Reduce the average time for operational implementations of new

satellite technology from two years to one

Accelerate use of research and operational satellite data in operational numerical prediction models

Goals:

52

Organizational Structure NASA NOAA DOD

Joint Oversight Board of Directors:NOAA NCEP: L. Uccellini (Chair)

Goddard ESD : F. EinaudiNOAA ORA: M. ColtonNOAA OAR: J. Gaynor

Navy: S. Chang, R. McCoyUSAF: J. Lanicci, M. Farrar

Joint Center StaffCenter Director: John LeMarshall

Executive Directors: Stephen Lord - NWSFuzhong Weng - NESDIS

L. P. Riishogjaard – NASAPat Phoebus - NRLTechnical Liaisons:

DAO – D. DeeEMC – J. Derber

GMAO – M. RieneckerOAR – A. GasiewskiORA – D. TarpleyNavy – N. Baker

USAF – M. McAteeProgram Support: Ada Armstrong

George Ohring (NESDIS)

AdvisoryPanel

Rotating Chair

ScienceSteering

Committee

53

JCSDA Partners

NASA/Goddard

Global Modeling & Assimilation Office

NOAA/NESDIS

Office of Research &

Applications

NOAA/OAR

Office of Weather and Air Quality

NOAA/NCEP

Environmental

Modeling Center

US Navy

Oceanographer of the Navy,Office of Naval Research (NRL)

US Air Force

AF Director of WeatherAF Weather Agency

PARTNERS

54

JCSDA Road Map (2002 - 2010)

Improved JCSDA data assimilation science

2002 2004 2007 2008 2009 2005

OK

Deficiency

2003

Advanced JCSDA community-based radiative transfer model,Advanced data thinning techniques

Sci

ence

Ad

van

ce

By 2010, a numerical weather prediction community will be empowered to effectively assimilate increasing amounts of

advanced satellite observations

2010

AMSU, HIRS, SSM/I, Quikscat,

AVHRR, TMI, GOES assimilated

AIRS, ATMS, CrIS, VIIRS, IASI, SSM/IS, AMSR, more products assimilated

Pre-JCSDA data assimilation science

Radiative transfer model, OPTRAN, ocean microwave emissivity, microwave land emissivity model, and GFS data assimilation system were developed

The radiances of satellite sounding channels were assimilated into EMC global model under only clear atmospheric conditions. Some satellite surface products (SST, GVI and snow cover, wind) were used in EMC models

A beta version of JCSDA community-based radiative transfer model (CRTM) transfer model will be developed, including non-raining clouds, snow and sea ice surface conditions

The radiances from advanced sounders will be used. Cloudy radiances will be tested under rain-free atmospheres, and more products (ozone, water vapor winds) are assimilated

NPOESS sensors ( CMIS, ATMS…) GOES-R

The CRTM includes scattering & polarization from cloud, precip and surface

The radiances can be assimilated under all conditions with the state-of-the science NWP models

Resources:

55

JCSDA FY03-04 Major Projects

• JCSDA is funding 18 extramural research projects to develop the state of-the-art satellite data assimilation system (e.g. uses of cloudy radiances from advanced satellite instruments, uses of satellite snow and vegetation products)

• Preparation for uses of advanced satellite data such as METOP (IASI/AMSU/HSB), DMSP (SSM/IS) and EOS (Aqua AIRS, AMSR-E)

• NCEP global data assimilation system implemented into NASA Global Modeling and Assimilation Office (GMAO) forecast system

• JCSDA community-based radiative transfer model

• Snow and sea ice emissivity models for improving uses of satellite microwave sounding data over high latitudes

• Impact studies of POES AMSU, EOS AIRS, DMSP SSMIS, MODIS, GPS Occultation on NWP through EMC parallel experiments

56

Recent Accomplishments Land emissivity model tested in NCEP operational models

Positive impacts with AMSU data over land (May,2002) Operational implementation (October, 2002)

Enabled use of microwave radiances over land

New Data used in NCEP operational models SSM/I, TRMM microwave imager precipitation estimates SSM/I, AMSU cloud liquid water GOES-10 IR radiances QuikSCAT data ( ~ 5 to 15% improvement in 10 m winds)

AIRS Assessment AIRS data examined at several intervals over the course of one year Full resolution tests indicate neutral impact in most forecast skill measures,

slight positive impact in others, focused on 254 channels out of 2378 Full data assimilation implementation scheduled for 1st Quarter FY05

Air Quality Prediction at NCEP

• Initial (1-5 years started FY2003) :

–1-day forecasts of surface ozone (O3) concentration

–Develop and validate in Northeastern US in 2 years

–Deploy Nationwide within 5 years

•Intermediate (5-7 years):–Develop and test capability to forecast particulate matter (PM) concentration

•Longer range (within 10 years):–Extend air quality forecast range to 48-72 hours

–Include broader range of significant pollutants

•Program has purchased additional computer power to perform AQF and promised this increment for perpetuity -92 -90 -88 -86 -84 -82 -80 -78 -76 -74 -72 -70 -68

34

36

38

40

42

44

46

48

0 to 10 1 0 to 20 2 0 to 30 3 0 to 40 4 0 to 50

Spatial Evaluation vs ObsHourly Ozone (rms srror)