Climate Prediction Modeling at NCEP

“Where America’s Climate, Weather, Ocean and Space Weather Services Begin”1

Dr. Louis W. UccelliniNational Centers for Environmental Prediction

Director

Committee on A National Strategy for Advancing Climate ModelingWashington, D.C.

October 3, 2011

2

Outline

• Define NCEP – Strategic Basis• Model Production Suite

– Link between weather and climate models• NCEP’s Climate Forecast System

– Recent upgrade to CFS Version 2.0– Reanalysis/reforecast

• Outreach– User/service workshop– COLA agreement– India MOU; model delivery

• Future CFS– Version 3.0 – engaging the community through the Climate Test

Bed– Multi-model ensemble

• Summary

NCEP Supports the NOAA Seamless Suite of Climate, Weather, and Ocean Products

Mission: NCEP delivers science-based environmental predictions to the nation and the global community. We collaborate with partners and customers to produce reliable, timely, and accurate analyses, guidance, forecasts, and warnings for the protection of life and property and the enhancement of the national economy.

3

Organization: Central component of NOAA National Weather Service

Vision: The Nation’s trusted source, first alert, and preferred partner for environmental prediction services

Space Weather Prediction Center NCEP Central Operations

Climate Prediction Center Environmental Modeling Center Hydromet Prediction CenterOcean Prediction Center

National Hurricane Center Storm Prediction Center

Aviation Weather Center

4

NOAA Seamless Suite of ForecastProducts Spanning Climate and Weather

Climate/WeatherLinkage

Week 2 Hazards Assessment

Forecast Forecast UncertaintyUncertainty

Forecast Forecast UncertaintyUncertainty

MinutesMinutes

HoursHours

DaysDays

1 Week1 Week

2 Week2 Week

MonthsMonths

SeasonsSeasons

YearsYears

CPC

Fore

cast

Lea

d Ti

me

Fore

cast

Lea

d Ti

me

Warnings & Alert Warnings & Alert CoordinationCoordination

WatchesWatches

ForecastsForecasts

Threats Assessments

GuidanceGuidance

OutlookOutlook

Benefits

TPCOPCHPC

SWPCAWCSPC

Service Center Perspective

Winter Weather Desk Days 1-3Tropical Storms to Day 5Severe Weather to Day 8

Fire Weather Outlooks to Day 8 :

NDFD, Days 4 -7

6-10 Day ForecastM

ariti

me

Mar

itim

eM

ariti

me

Mar

itim

e

Life

& P

rope

rty

Life

& P

rope

rty

Life

& P

rope

rty

Life

& P

rope

rty

Spac

e O

pera

tions

Spac

e O

pera

tions

Spac

e O

pera

tions

Spac

e O

pera

tions

Recr

eatio

nRe

crea

tion

Recr

eatio

nRe

crea

tion

Ecos

yste

mEc

osys

tem

Ecos

yste

mEc

osys

tem

Envi

ronm

ent

Envi

ronm

ent

Envi

ronm

ent

Envi

ronm

ent

Emer

genc

y M

gmt

Emer

genc

y M

gmt

Emer

genc

y M

gmt

Emer

genc

y M

gmt

Agric

ultu

reAg

ricul

ture

Agric

ultu

reAg

ricul

ture

Rese

rvoi

r Con

trol

Rese

rvoi

r Con

trol

Rese

rvoi

r Con

trol

Rese

rvoi

r Con

trol

Ener

gy P

lann

ing

Ener

gy P

lann

ing

Ener

gy P

lann

ing

Ener

gy P

lann

ing

Com

mer

ceCo

mm

erce

Com

mer

ceCo

mm

erce

Hyd

ropo

wer

Hyd

ropo

wer

Hyd

ropo

wer

Hyd

ropo

wer

Fire

Wea

ther

Fire

Wea

ther

Fire

Wea

ther

Fire

Wea

ther

Hea

lthH

ealth

Hea

lthH

ealth

Avia

tion

Avia

tion

Avia

tion

Avia

tion

Seasonal Predictions

Forecast Forecast UncertaintyUncertaintyForecast Forecast UncertaintyUncertainty

MinutesMinutes

HoursHours

DaysDays

1 Week1 Week

2 Week2 Week

MonthsMonths

SeasonsSeasons

YearsYears

NWS Seamless Suite of ForecastProducts Spanning Weather and Climate

Fo

reca

st

Lea

d T

ime

Fo

reca

st

Lea

d T

ime

Warnings & Alert Warnings & Alert CoordinationCoordination

WatchesWatches

ForecastsForecasts

Threats Assessments

GuidanceGuidance

OutlookOutlook

Benefits

NCEP Model Perspective

Mar

itim

e

Mar

itim

e

Life

& P

rope

rty

Life

& P

rope

rty

Spa

ce O

pera

tions

Spa

ce O

pera

tions

Rec

reat

ion

Rec

reat

ion

Eco

syst

em

Eco

syst

em

Env

iron

men

t

Env

iron

men

t

Em

erge

ncy

Mgm

t

Em

erge

ncy

Mgm

t

Agr

icul

ture

Agr

icul

ture

Res

ervo

ir C

ontr

ol

Res

ervo

ir C

ontr

ol

Ene

rgy

Pla

nnin

g

Ene

rgy

Pla

nnin

g

Com

mer

ce

Com

mer

ce

Hyd

ropo

wer

Hyd

ropo

wer

Fire

Wea

ther

Fire

Wea

ther

Hea

lthH

ealth

Avi

atio

n

Avi

atio

n

•North American Ensemble Forecast System

•Climate Forecast System (EUROSIP*)

•Short-Range Ensemble Forecast

•Global Forecast System

•North American Mesoscale

•Rapid Update Cycle for Aviation•Dispersion Models for DHS

•Global Ensemble Forecast System

Hurricane WRF & GFDL

WavesReal Time Ocean Forecast System

Space Weather Tsunami

*To become available for NCEP operational seasonal prediction in Dec 2011

• EMC WRF Developmental Test Center, Joint Center for Satellite Data Assimilation

• CPC Climate Test Bed

• NHC Joint Hurricane Test Bed

• HPC Hydrometeorological Test Bed

• SPC Hazardous Weather Test Bed with NSSL

• SWPC Space Weather Prediction Test Bed with AFWA

• AWC Aviation Weather Test Bed

• OPC IOOS Supported Test Bed (in discussion with NOS/IOOS)

Test BedsService – Science Linkage with the Outside Community:

Accelerating the R2O Transition Process

6

7

Launch List – Model Implementation Process

EMCNCO

R&D Operations Delivery

Criteria

Transition from Research to Operations (R2O)

Requirements

EMC

Schematics in the Model Transition Process

OPS Life cycleSupport

Service Centers

NOAAResearch

Concept of Operations

ServiceCenters

User

Ob

serv

atio

n

Sys

tem

FieldOffices

Effort

EMC and NCO have critical roles in the transition from NOAA R&D to operations

Other Agencies&

International

Forecast benefits, Efficiency, IT Compatibility, Sustainability

ASI, COLA,ARCS

Operations to Research (O2R)

Test BedsCTB

JCSDAWRF DTC

JHT . .

“To Accelerate R2O,Need to Support O2R”

8

Model Production SuiteLink between weather and climate models

9

Air Quality

WRF NMM/ARWWorkstation WRF

WRF: ARW, NMMETA, RSM GFS, Canadian Global Model

Regional NAMWRF NMM

North American Ensemble Forecast System

Hurricane GFDLHWRF

GlobalForecastSystem

Dispersion

ARL/HYSPLIT

Forecast

Severe Weather

Rapid Updatefor Aviation

Climate ForecastSystem

Short-RangeEnsemble Forecast

NOAA’s Model Production Suite

GFS MOM4NOAH Sea Ice

NOAH Land Surface Model

Coupled

Global DataAssimilation

OceansHYCOM

WaveWatch III

NAM/CMAQ

9

Regi

onal

DA

Satellites + Radar99.9%

~2B Obs/Day

NOS – OFSGreat Lakes

Bays• Chesapeake• Tampa • Delaware

SpaceWeather

(Future)

ENLIL

Regi

onal

DA

10

NCEP’s Climate Forecast System(CFS)

11

NCEP Requirement for CFS

• CFS is the first dynamic model used for NCEP operational intra-seasonal and inter-annual forecasts– Operated in an ensemble mode (4/day)– To be combined with EUROSIP models as part of an International Multi-

model Ensemble (by December 2011)• Provides real-time ENSO forecast for El Nino, La Nina alerts,

watches and warnings• The CFS-based coupled Reanalysis

provides the best estimate of the state of the coupled climate system, which is the basis for operational climate monitoring and analyses.

• The CFS-based Reforecast provides basis for model calibration of CFS real time forecast used in CPC operations.

–

CFSv1 implemented

CTB spun up

CFSv2 implemented

48 month running mean of Heidke Skill scores computed for seasonal outlooks of U.S. Surface Temperature for each 3-month seasonal mean

US Seasonal Temperature GPRA* Measure

*GPRA – Government Performance and Results Act

13

Attribute CFS v1.0

Operational Since 2004

CFS v2.0

Operational Since March 2011Analysis Resolution 200 km 27 km (T574)

Atmosphere model 2003: 200 km/64 levels

Humidity based clouds

2010: 100 km (T126)/64 levels

Variable CO2

AER SW & LW radiation

Prognostic clouds & liquid water

Retuned mountain blocking

Convective gravity wave drag

Ocean model MOM-3: 60N-65S

1/3 x 1 deg.

Assim depth 750 m

MOM-4 fully global

¼ x ½ deg.

Assim depth 4737 m

Land surface model (LSM) and assimilation

2-level OSU LSM

No separate land data assim

4 level Noah model

GLDAS driven by obs precip

Sea ice Climatology Daily analysis and 3-layer interactive sea ice model

Coupling Daily 30 minutes

Data assimilation Retrieved soundings, 1995 analysis, uncoupled background

Radiances assimilated, 2008 GSI, coupled background

Reforecasts

(Real-time extension)

15/month seasonal output 24/month (seasonal)

124/month (week 3-6)

NOAA Climate Forecast System (CFS) – Recent upgrade (March 30, 2011)

T126

Climate Forecast System Reanalysis (CFSR)

• A T384/L64 (~ 38 km) reanalysis from 1979 – present (GFS/GSI – Atmosphere; MOM4 -Ocean; NOAH/GLDAS – Land; Sea Ice model)

• A semi-coupled data assimilation– 6 hour guess forecast from a coupled model– Followed by independent ocean and atmosphere reanalysis

• Implemented operationally in March 2011• Data availability from National Climatic Data Center

CFSR data dumps (GB/Month)

Number of OceanTemperature Profiles/month 1980-2009

Time

<--

Dep

th

Climate Reforecast for CFS Version 2

• A coupled prediction system for extended-range and seasonal predictions ; Implemented March 2011, – Atmospheric model -resolution T126, 64 vertical levels– Ocean model (MOM4) horizontal resolution: 1/2 Deg. in zonal direction; 1/4

Deg between 10S-10N gradually increasing to 1/2 Deg poleward of 30S and 30N, Vertical Resolution: 40 layers; with 27-layers in upper 400m, and a bottom at approximately at 4.5 km in the ocean

• Atmosphere/Ocean/Land/Sea Ice Initial conditions from the CFS Reanalysis

• Reforecasts for calibration– Seasonal (9-month): 1981 – 2010 (4 runs every five days)– Extended-range (45-day) – 1999-2010 (4 runs every day)– Over 10,000 years of reforecasts

• Data availability from the NCDC

16

Outreach

• CFS Needs Assessment Workshop (Mar 2011)

• COLA agreement (Jul 2009)• India MOU (Nov 2010; model delivery

(Apr 2011)

• Goals– Identify CFS users’ needs for model codes and datasets– Develop formal prioritized needs assessment– Develop priority for real-time availability– Identify gaps– Scope user needs for NMME

• Workshop attended by 50 (private industry, academia, federal, state and local agencies)

• Results/Requirements– Internal services agreement between NCEP and NCDC in

development to identify issues, requirements and options for high priority CFS development, execution, archive and dissemination activities

• Real-time and reliable access to CFS ensembles, reanalysis, reforecast• Refinements to the reanalysis including downscaling, and at model

resolution• More open process for defining CFS version 3• Reanalysis Lite (unify resolution)

17

CFS Needs Assessment WorkshopMarch 8, 2011

COLA-NCEP CFS Collaboration

• Letter of agreement: “Access to the NCEP Climate Forecast System” signed 15 July 2009

– Over 40 peer-reviewed papers 2008 - 2011

• Examples of CFS.v2 research efforts at COLA– Decadal experiments (1960-2010 initial conditions) per CMIP5 protocol– Extension of CFS.v2 to the CFS Interactive Ensemble (CFSIE)– Long free simulations with CFS.v2 and their analysis– Hypothesis-based experiments to address the weak AMOC variability in CFS.v2– Diagnostics of CFS.v2 hindcasts

• COLA ported CFS.v2 on different computing platforms (with some help from NCEP); COLA’s experience will help NCEP port CFS.v2 on NOAA’s research HPC

Weaker AMOC Drift in Sea Ice thickness – Too much

melt?

Implementation Agreement on Modeling Between NCEP and Ministry of Earth Sciences (MoES), India

• Implementation agreement signed November, 2010– To develop and transfer extended range and seasonal forecast

systems that will permit more accurate and timely predictions of Indian Monsoons for enhancing food security and other agricultural uses

• April 2011 - NCEP scientists delivered the operational version of the GFS/GSI and CFS.v2 to the MoES

• April 11-15, 2011, NCEP scientists held a workshop at the Indian Institute of Tropical Meteorology (IITM) on specialized training on the GFS/GSI and CFS.v2

• A jointly funded "Monsoon Desk" has been set up at NCEP to coordinate numerical model simulations and diagnostics with IITM and IMD during the next five years

• Already documenting improved Monsoon, precipitation and MJO forecasts

20

Future CFS– Version 3.0 – engaging the community through the Climate

Test Bed– Multi-model ensemble opportunities

• Organized by Climate Test Bed (CTB) • 48 participants: scientists from NCEP (EMC and CPC), NOAA management,

US modeling centers (GFDL, NCAR, NASA/GSFC), international modeling centers (ECMWF, Italy/CMCC, India/IITM, Taiwan/CMB), other centers/labs of NOAA (PMEL, NWS/OST, ESRL, NWS/OHD) and scientists from universities and non-profit research institutes (COLA).

• Define NOAA’s requirements and strategy for CFSv3• Role of CTB to enhance NCEP collaborations with external community for CFS

development• Identify range/role of CFS within MMEs (international and national): MME’s

representing a major “force for change” in the model and service provider communities

• Next Steps• Hold a CFS Science Workshop focused on CFSv2 evaluations in Spring, 2012• Develop a White Paper on CFSv3 Development Strategy based on recommendations

from the CFSv3 planning meeting

Expectations

Results of CFS Version 3 Planning Meeting August 25-26, 2011

22

Commitments for CFSv3

• NCEP Human Resources• EMC:

o Model implementation o Data assimilationo Reanalysis/reforecast

• NCOo Model implementation o Sustaining the operational model suite

• CPC: o Model diagnosis/evaluationo Applications for operational ISI forecasts

• CTB: o Outreach to the research communityo Providing a model testing environment for CFS evaluation, diagnosis and improvement o Supporting Climate Process Teamso Facilitates the transfer of specific advances back into NCEP operational version

• NOAA Computer Resourceso GAEA – Site Ao ZEUS – Site Bo Operational WCOSS

23

Future

• IMME – EUROSIP• NMME

– working toward “experimental real-time”– Coordinated and assessed through the Climate Test Bed– Working toward a CONOPS for operational real-time

prediction

• Real-time archive CFS– Reanalysis– Reforecast– Daily forecasts

Table 1. NMME Phased Strategy – coordinated through the Climate Test Bed

Activities Description Forecast Type

Participating Models

Where, What and Who

Financial Support

Data Archive and Distribution

Phase-I: NMME Opportunity

Ad hoc NMME system based on existing hindcasts from each modeling center

Experimental NMME forecastStarted in August 2011

NCEP (CFSv1, CFSv2)GFDL (CM2.2)IRI (Echam4-a & 4-f)NCAR (CCSM3.0)NASA (GEOS5)

i) Each modeling center runs the reforecasts and real-time forecast in their own computers. ii) NCEP obtains the hindcasts and real time forecasts and makes routine real-time monthly and S/I forecasts

CPO i) Each modeling center provides data to the NMME team members; ii) IRI disseminates the monthly mean hindcast and real-time data (3 variables)

Phase-II:Purposeful NMME

A designed NMME system as a research prototype in preparation for operations

Experimental NMME forecast during the FY12-13 research project

Potential models:NCEP (CFSv2)GFDL (CM2.5)IRI (Echam4)NCAR (CCSM4.0, and CESM1.0) NASA (GEOS5)

Same as Phase-I, except NCAR/CCSM4 and ICHAM models will generate their hindcasts and real time forecasts at NOAA computer (GAEA)

CPO i) Each modeling center provides data to the NMME team members; ii) IRI archive the monthly data; iii) and NCDC archives hindcasts and realtime data including daily data from selected models (resources required)

Phase-III:Operational NMME

An operational NMME ISI prediction system at NCEP

Operational NMME Forecast after FY13 and after the value of NMME is demonstrated

TBD Three options are listed in Table 2. The issue needs to be resolved through an open review process. For all options, NCEP will obtain the data and make operational real time ISI forecasts.

CPO NCDC archives and disseminates the NMME hindcast and real time data (resources required)

Table2. Three Options for Hindcasts and Real time Forecasts in the Operational NMME Prediction System

Operational NMME

Reforecasts Real-time forecasts

Pros Cons and Requirements

Option 1 Each modeling center Each modeling center

1) Easy to be transitioned to operations since it’s same as Phase-I and II; 2) minimum workload for NCEP (i.e., NCEP needs only to obtain hindcasts and real-time data and make real time forecasts); 3) easy to upgrade models; 4) minimum NCO involvement; 5) can leverage other modeling centers efforts since they general hindcasts for their own uses (e.g., model evaluations);

Need to sign agreements between NCEP and each modeling center to ensure real time delivery of hindcasts and real time forecasts. The agreements should include required resources (financial, HPC, personnel).

Option 2 NCEP NCEP Delivery of real-time operational forecasts is guaranteed

1) NCEP needs to port other models to NCEP computer; 2) NCEP needs to train personnel to run other models to generate hindcasts and real-time forecasts; 3) requires tremendous storage; 4) hard to upgrade models

Option 3 Each modeling center NCEP 1) Leverage other modeling centers efforts since they generate hindcasts for their own uses (e.g., model evaluations and calibration) and thus less work for NCEP; and 2) less storage requirement at NCEP compared to Option 2.

1) NCEP needs to port other models to NCEP computer; 2) NCEP needs to train personnel to run other models to make real-time forecasts; and 3) hard to upgrade models, 4) Need to sign agreements between NCEP and each modeling center to ensure on-time delivery of hindcasts.

26

Summary

• NCEP has successfully provided an operational Climate Forecast System– First dynamical model used by CPC– Improved seasonal/ENSO forecasts

• Research community has become increasingly involved with the use of CFS version 1 and 2 for monsoon, MJO, decadal, and other research topics

• NCEP is committed to support version 2 to version 3 improvements and related “O2R” “R2O” paradigm

• NCEP is positioned to help address ongoing opportunities/challenges related to creating a real-time NMME

27

Appendix

Computing Capability

• Transition to IBM Power 6 complete– Declared operational

August 12, 2009– 73.1 trillion calculations/sec– Factor of 4 increase over the

IBM Power5 – 156 POWER6 32-way nodes– 4,992 processors– 20 terabytes of memory – 330 terabytes of disk space– 3.5 billion observations/day– 27.8 million model fields/day

• Primary: Gaithersburg, MD• Backup: Fairmont, WV

– Guaranteed switchover in 15 minutes

– Web access to models as they run on the CCS

“reliable, timely and accurate”

0

10

20

30

40

50

60

70

80

90

JULJANJULJANJULJANJULJANJULJANJULJANJULJANJULJANJULJANJULJANJUL

Num

ber o

f Hits

(Mill

ions

)2001 2002 2003 2004 2005 2006 2007 2008 2009

Popularity of NCEP Models Web Page

2010