PROWESS 2014 1

Dr. William. M. Lapenta Director

National Centers for Environmental Prediction

NOAA/NWS

J. C. Derber, G. DiMego, M. B. Ek, M. Iredell, S. Moorthi, V. Tallapragada,

H. L. Tolman, S. Saha, and Y. Zhu

NCEP

Operational Numerical Guidance for

NOAA Weather and Climate Services

PROWESS 2014

NOAA Operational Numerical Guidance

Supports the Agency Mission

– Numerical Weather Prediction at NOAA

Required for agency to meet service-based metrics

– National Weather Service GPRA* Metrics (* Government Performance & Results Act)

Hurricane Track and Intensity

Winter Storm Warning

Precipitation Threat

Flood Warning

Marine Wind Speed and Wave Height

– Operational numerical guidance: Foundational tools used by government, public and private

industry to improve public safety, quality of life and make

business decisions that drive U.S. economic growth

2

Lead Time

and

Accuracy!

PROWESS 2014

Forecast

Uncertainty

Minutes

Hours

Days

1 Week

2 Week

Months

Seasons

Years

Seamless Suite of Operational

Numerical Guidance Systems

Fo

rec

as

t L

ea

d T

ime

Warnings & Alert

Coordination

Watches

Forecasts

Threats

Assessments

Guidance

Outlook

Benefits

•North American Ensemble Forecast System •Climate Forecast System

•Short-Range Ensemble Forecast

•Global Forecast System

•North American Mesoscale

•Rapid Refresh

•Dispersion (smoke)

•Global Ensemble Forecast System

• Regional Hurricane • (HWRF & GFDL)

• Waves • Global Ocean

• Space Weather

Spanning Weather and Climate

• Tsunami

• Whole

Atmosphere

• HRRR

• NMME

• NLDAS

• Wave Ensemble

• Bays

• Storm Surge

•Global Dust

•Fire Wx

3

• Air Quality

PROWESS 2014

Regional Hurricane

GFDL WRF-NMM

WRF(ARW, NMM) NMMB

Climate Forecast System (CFS)

Short-Range Ensemble Forecast

NOAA’s Operational Numerical

Guidance Suite (Jan 2014)

GFS, MOM4, NOAH, Sea Ice

North American Ensemble Forecast System

GEFS, Canadian Global Model

Dispersion HYSPLIT

Air Quality

CMAQ

Regional NAM

NMMB NOAH

3D

-VA

R

DA

Regional Bays •Great Lakes (POM)

•N Gulf of Mexico (FVCOM) •Columbia R. (SELFE) •Chesapeake (ROMS)

•Tampa (ROMS) •Delaware (ROMS)

Space Weather

ENLIL 4

North American Land Surface Data Assimilation

System NOAH Land Surface Model

Global Spectral NOAH

3D

-En

-Var

D

A

Global Forecast System (GFS)

3D

-VA

R

DA

3D

-VA

R

DA

WRF ARW

Rapid Refresh

3D

-VA

R

DA

Waves WaveWatch III

Ocean HYCOM

Ecosystem EwE

Global Ensemble Forecast System (GEFS)

21 GFS Members

ESTOFS ADCIRC

SURGE SLOSH

P-SURGE SLOSH

WRF ARW

3D

-VA

R

DA

High Resolution RR

NEMS Aerosol Global Component (NGAC)

GFS & GOCART

WRF(ARW, NMM) & NMMB

High Res Windows

PROWESS 2014 5

24-h Cycle 01 February 2014

Numerical Guidance Suite Execution

WCOSS NOAA Supercomputer

Nu

mb

er

of

No

de

s

Time of Day (UTC)

00 06 12 18 00 CFS NAM GFS GEFS SREF

HyCOM

SREF

Para RAPv2

Para

RAP

PROWESS 2014 6

Forcing Factors Shaping NOAA

Operational Numerical Guidance

Emerging Requirements Weather Ready Nation

High impact events

Weather to climate—seamless suite of guidance and products

Science and Technology Advances Observing systems

High performance computing

Data dissemination

Numerical Guidance Systems Data assimilation (methodology)

Modeling (physics, coupling & dynamics)

Ensembles (construction—initialization, membership, etc.)

Intelligent post processing

Predictability convective systems

Seasonal to interannual

PROWESS 2014 7

The NOAA Modeling Strategy…

High Level Perspective Moving away from the “model of the day”

Continue to pursue multi-model approach to ensembles

• Don’t forget: ensemble systems only as good as the modeling

system it is built from

Priorities for deterministic development are clear:

1. Data assimilation (methodology and observations)

2. Model physics

Why do we continue to underplay this important part of the

enterprise?

Clouds, microphysics, radiation, land, ocean, ice,

aerosols….includes coupling

3. Resolution—horizontal and vertical

4. Dynamic core

Must consider advanced HPC technologies but don’t forget about

the science

Regional systems shift to convection permitting applications

NOAA encouraged to consider unified modeling approach

(UCACN)

PROWESS 2014

System Current Q4FY14 FY18

GDAS

80 member @ T254

Eulerian (55 km)

80 member @ T574 SL (35

km)

4DHybrid 80 member @

T1148 SL (17 km)

Analysis @ T574

Eulerian (27 km) using

T254 Eulerian ensembles

Analysis Increment @ T574

SL (27 km) using T574 SL

ensembles

Analysis Increment @T1148

SL (17 km) using T1148 SL

ensembles

64 Vertical Levels 64 Vertical Levels 128 Vertical Levels

Additional Obs., Improved

radiative transfer, many

smaller changes, uses GFS

model below

Additional Obs., Cloudy

Radiances, Improved QC and

ob. Errors, Ensemble

Hurricane relocation

System Current Q4FY14 FY18

GFS

T574 Eulerian (27 km) to

7.5 days

T1534 SemiLagrangian (13

km) to 10 days

T2000 (10 km) to 10 days

T254 Eulerian (55 km)

days 7.5 to 16

T574 SL (35 km) days 10 to

16

T1148 (17 km) days 10 to 16

64 Vertical Levels 64 Vertical Levels 128 Vertical Levels

Global Forecast System Evolution: 2014 to 2018

PROWESS 2014

Real Time Ocean Forecast System Implemented 24 October 2011

9

• RTOFS Global is the first global eddy-resolving ocean forecast system at NOAA/NCEP

• 1/12 degree HYCOM (HYbrid Coordinate Ocean Model)

• 6-day forecasts initialized at 00 UTC

• 32 vertical hybrid layers (isopycnal in the deep, isolevel in the mixed layer and sigma in shallow waters)

• Initialization: MVOI scheme (NCODA) developed by the US Navy

• Forced with the GFS surface fluxes of radiation, precipitation and momentum.

Thanks to the NAVY for

partnering with NOAA and

making HYCOM available

PROWESS 2014

Operational Wave Model

Configurations

10

Multi-1 global model.

Multi-2 global hurricane model.

Great lakes with 2 staggered cycles

(NAM driven and NDFD driven)

NCEP wave ensemble system

Combined FNMOC/NCEP wave

ensemble system

Nearshore Wave Prediction System

(Alpha testing)

PROWESS 2014 11

• 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 Institute for Climate and Society (IRI)

• Canadian Meteorological Centre (CMC) • NASA – GMAO

• NOAA/NCEP/EMC/CPC

• NOAA/GFDL

• Princeton University

• University of Colorado (CIRES

National Multi-Model

Ensemble (NMME) Project

Data are available at: http://iridl.ldeo.columbia.edu/SOURCES/.Models/.NMME/

PROWESS 2014 12

Model Hindcast

Period

No. of

Member

Arrangement of

Members

Lead

(months)

Model

Resolution:

Atmosphere

Model

Resolution:

Ocean

Reference

NCEP-CFSv2 1982-2010 24(20)

4 members

(0,6,12,18Z)

every 5th day

0-9 T126L64 MOM4 L40

0.25 deg Eq

Saha et al.

(2010)

GFDL-CM2.1 1982-2010 10 All 1st of the month

0Z 0-11 2x2.5deg L24

MOM4 L50

0.30 deg Eq

Delworth et

al. (2006)

CMC1-

CanCM3 1981-2010 10

All 1st of the month

0Z 0-11

CanAM3

T63L31

CanOM4 L40

0.94 deg Eq

Merryfield

et al. (2012)

CMC2-

CanCM4 1981-2010 10

All 1st of the month

0Z 0-11

CanAM4

T63L35

CanOM4 L40

0.94 deg Eq

Merryfield

et al. (2012)

NCAR-

CCSM3.0 1982-2010 6 All 1st of the month 0-11 T85L26

POP L40 0.3

deg Eq

Kirtman and

Min (2009)

NASA-

GEOS5 1981-2010 11

4 members every 5th

days; 7 members on

the last day of the

previous month

0-9 1x1.25deg L72 MOM4 L40

1/4 deg at Eq

Rienecker et

al. (2008)

* This slide is by courtesy of Huug Vandendool, Qin Zhang, and Emily Becker.

Current NMME Models

Benefits of NMME

to Contributing Agencies • A real-time prediction platform for model diagnosis and

evaluations of their perspective model(s) with comparison of

their models for both weather and monthly/seasonal time scales

in an operational setting

• Enhanced coordination and collaboration on climate

predictability and prediction studies including guidance for

their model evaluation and development efforts;

• A platform to demonstrate broader impacts of their supported

research advances on decision makers and end-users who

depend on climate information at ISI time scales, and

• Publicly available NMME data for research and users

community including young scientists.

PROWESS 2014 14

NAM • Implemented 18 October 2011

• NEMS based NMM

• Parent remains at 12 km

• Multiple Nests Run to ~48hr – ~4 km CONUS nest

– ~6 km Alaska nest

– ~3 km HI & PR nests

– ~1.5-2km DHS/FireWeather/IMET possible

Rapid Refresh • Scheduled implementation 20 March 2012

• WRF-based ARW

• Use of GSI analysis

• Expanded 13 km Domain to include Alaska

• Experimental 3 km HRRR

RUC-13 CONUS domain

WRF-Rapid Refresh domain – 2010

Original CONUS domain

Experimental 3 km HRRR

NCEP Mesoscale Modeling

for CONUS:

Only 14 cases at 120 hr

verified for 2013

NCEP Operational HWRF showing systematic improvements in intensity forecasts

Intensity Forecast Errors from Operational HWRF 2008-2013

15

16

Advancements to Operational HWRF – Transition to NMM-B/NEMS Multi-Scale Modeling System

• NCEP/AOML Collaborative effort supported by OAR

Sandy Supplemental High Impact Weather Prediction

Project (HIWPP) and leveraged by NOAA’s HFIP support

• Take advantage of NMMB in NEMS infrastructure for

developing next generation global-to-local-scale

modeling system for tropical cyclone forecasting needs

and for comprehensive solutions for landfalling storms

• Planned development, testing and evaluation leading to

potential transition to operations in the next 3-5 years

Scientific advancements include:

• Scale aware and feature aware

physics for high-resolution

domains and for multi-scale

interactions

• Advanced techniques for inner

core data assimilation with use all

available aircraft recon data

including TDR, FL, SFMR, and

satellite radiance data

• High-resolution ensembles for

prediction of RI/RW

• Enhanced land-air-sea-wave-

hydrology coupled system

16

PROWESS 2014 17

• The GFS must: Provide skillful guidance for all NOAA operational applications

Demonstrate at least the computational efficiency of the current GFS

Provide the flexibility to meet future demands

• Candidate Models Currently Under Development

GFS Global Spectral Model (GSM--NCEP)

Non-Hydrostatic Multiscale Model (NMM--NCEP)

Non-hydrostatic Flow Following Icosahedral Model (NIM--ESRL)

Cubed-Sphere Finite Volume (HiRAM--GFDL)

Model for Prediction Across Scales (MPAS--NCAR)

• NCEP Committed to global spectral model through 2018

Viable candidate MUST emerge with acceptable level of risk to opreations

Cycles with data assimilation system and robust physics package

Computationally efficient (conventional and advanced architectures)

Applicable to medium range and ISI

Next Operational Global Model?

•Two racks of hybrid x86 / GPU nodes

•96 nodes per rack

•Each node contains: • 16 core AMD Opteron processor

• 1 NVIDIA Kepler GK110 module

• Each Kepler GPU module contains: • 1.0+ Teraflops – Double precision precision floating point performance

• 14 SMX units (192 single-precision CUDA cores / SMX unit)

• Totals across two racks: • ~200+ Teraflops

• 2,944 AMD Opteron cores

• 184 GPU / 2,576 SXM Units / 494,592 CUDA cores

• Separate system integrated into existing Gaea infrastructure

• First customers of Cray XK6 technology • Methods for measuring and allocating time will evolve over time

• Development teams need to be proactive in reporting problems

Balance Between Science

and Technology Required

18

PROWESS 2014 19

Thanks for Your Time.

Questions?