Global ClimateGlobal ClimateModelingModeling

Warren M. WashingtonNCARApril 2000

Global COGlobal CO22 Trend and Rate of Increase Trend and Rate of Increase

Global Methane TrendsGlobal Methane Trends

Snow Cover TrendsSnow Cover Trends

Greenland Ice ThinningGreenland Ice Thinningu Melting in low-lying areas is very rapid

at about 3 feet per year….ScienceMagazine

u Most melting on the southern andeastern parts of Greenland

u Snowfall not changed much…..it is theglaciers moving faster and melting

Expected Changes in theExpected Changes in theEarth’s ChemistryEarth’s Chemistry

u Increased carbon dioxide,methane, nitrogen oxide, loweratmosphere ozone

uDecreased CFCs

u Increased aerosols from fossil fuelcombustion, biomass burning

ParallelParallelClimate ModelClimate Model

(PCM)(PCM)

Laws of Physics inLaws of Physics inClimate ModelsClimate Models

u Conservation of Momentumu First Law of Thermodynamicsu Conservation of Massu Ideal Gas Lawu Hydrostatic Assumptionu Conservation of Water (Vapor and Liquid)u Some models have active atmospheric

chemistry and aerosol physics

PCM Component ModelsPCM Component ModelsuAtmosphere

– NCAR CCM3.2– T42 18 levels– land surface model embedded– SPMD option: 1-D data decomposition (64pe limit)

uOcean– Parallel Ocean Program (POP)– ~2/3 degree horizontal displaced polar grid– 32 levels– 2-D data decomposition

More ComponentsMore ComponentsMore Componentsu Sea ice-viscous-plastic Hibler dynamics with

relaxation, option for elastic-viscous-plastic, 27km resolution, and Parkinson-Washingtonthermodynamics (Zhang, Semtner-NPS;Weatherly-CRREL; Craig-NCAR; Hunke,Dukowicz-LANL)

u Parallel flux coupler (Craig, Bettge, Loft, Dennis,James-NCAR; Jones-LANL)

u River Transport Model (Branstetter, Famiglietti-U. Texas, Austin; Craig-NCAR)

Sequential Executionof PCM

PCM Flux Coupler

CCM3 SeaIce

RiverTransportPOP

ParallelParallelComputersComputers

PCM 1.1 has been run on the following distributedand shared memory systems:

è CRAY T3E900è SGI Origin 2000/128è HP SPP2000è IBM SP2è Sun Starfireè DEC/Compaq Alpha Clusterè Linux Cluster

PCM 1.1PCM 1.12/3o POP

27 km sea ice

T42 CCM3.2

View of the Parallel Ocean Program (POP)Model Horizontal Grid at 2/3o Resolution

Note highresolutionin NorthAtlanticand nearequator.

Numerical MethodsNumerical Methodsu Atmosphere- spherical harmonics, FFT,FD,

Lagrangian, semi-implicitu Ocean-FD, solution of Laplacian equation,

semi-implicit

u Sea Ice-FD

Sequential Executionof PCM

PCM Flux Coupler

CCM3 SeaIce

RiverTransportPOP

Moving from MessageMoving from MessageParadigm to DistributedParadigm to Distributed

Shared Memory ParadigmShared Memory Paradigm

MessageMessagePassingPassing

SharedSharedMemoryMemory

DistributedDistributedSharedSharedMemoryMemory

MPICommunicationbetween PEs

Open MPdirectives withinnode

Open MPdirectives withinnode andmessage passingbetween nodes

Global Atmosphere

Global OceanGlobal Ocean

Sea IceSea Ice

Deep/Abyssal Northwest AtlanticDeep/Abyssal Northwest Atlantic

Regional ClimateRegional ClimateAspectsAspects

u ENSOu Arctic Oscillation

u North AtlanticOscillation

u AntarcticCircumpolarWave

Examples of ClimateExamples of ClimateChange ExperimentsChange Experiments

u Greenhouse gases

u Sulfate aerosols (direct and indirect)

u Stratospheric ozone

u Biomass burning

u Historical simulations

u Various energy/emissions use strategies

Solar VariabilitySolar VariabilitySimulationsSimulations

u In addition to Greenhouse gasesand sulfate aerosol effects

uOne of the ensemble shows aglobal surface temperature changesimilar to the observed record

1860 1880 1900 1920 1940 1960 1980 2000

0.0

0.2

0.4

0.6

-0.2

deg

rees

C

Year

Global Surface Temperature Anomaly

Observed PCM Simulation

1860-1999

Present and Future Emphasisfor Climate Modeling

uMore detailed interactions ofatmosphere, land/vegetation/riverrunoff, ocean and sea ice

uUse of large parallel clusteredcomputer systems

uParadigm -- distant collaborations

ParallelParallelComputersComputers

PCM 1.1 has been run on the following distributedand shared memory systems:

è CRAY T3E900è SGI Origin 2000/128è HP SPP2000è IBM SP2è Sun Starfireè DEC/Compaq Alpha Clusterè Linux Cluster

PCM 1.1PCM 1.12/3o POP

27 km sea ice

T42 CCM3.2

DOE/UCAR ParallelClimate Model Project

IBM/SP O2K300

Need Help from AppliedNeed Help from AppliedMathematics CommunityMathematics Community

u Remappingu Laplacian solver on cluster computer systemsu Global sums, fast FFTsu Re-coding for distributed cluster computers

systemu Better use of cachesu Improved numerical methods for atmosphere,

ocean, sea ice, and hydrological system

Distributed InvolvementDistributed InvolvementDOE and NSF Supported Project withDOE and NSF Supported Project with::

uu Los Alamos National LaboratoryLos Alamos National Laboratoryuu National Center for Atmospheric ResearchNational Center for Atmospheric Researchuu Naval Postgraduate SchoolNaval Postgraduate Schooluu Oak Ridge National LaboratoryOak Ridge National Laboratoryuu University of Texas, AustinUniversity of Texas, Austinuu Scripps Oceanographic InstituteScripps Oceanographic Instituteuu DOE Program on Climate Diagnostics and IntercomparisonDOE Program on Climate Diagnostics and Intercomparisonuu U.S. Army Cold Regions Research and EngineeringU.S. Army Cold Regions Research and Engineering

LaboratoryLaboratoryuu National Energy Research Supercomputer CenterNational Energy Research Supercomputer Center

Animation CreditsAnimation Creditsu The atmospheric animation was from the

Community Climate Model at T170 resolution.This model was developed by the NCAR ClimateModeling Section. The graphics were preparedby Don Middleton of NCAR.

u The ocean animation makes use of the LANLPOP model and was prepared by the scientists atthe Naval Postgraduate School (NPS)

u The sea animation uses the Zhang model of theNPS.

The EndThe End

More information can be found athttp://www.cgd.ucar.edu/ccr/pcm/