Fine-resolution global time slice Fine-resolution global time slice simulationssimulations

Fine-resolution global time slice Fine-resolution global time slice simulationssimulations

Philip B. DuffyPhilip B. Duffy1,2,31,2,3

Collaborators: G. BalaCollaborators: G. Bala11, A. Mirin, A. Mirin11

11Lawrence Livermore National LaboratoryLawrence Livermore National Laboratory22University of California, MercedUniversity of California, Merced 22University of California, DavisUniversity of California, Davis

NARRCAP Users’ Meeting, February 14, 2008

THIS TALK APPROVED FOR

Why include global-domain Why include global-domain simulations in NARCCAP?simulations in NARCCAP?Why include global-domain Why include global-domain simulations in NARCCAP?simulations in NARCCAP?

• Nice to have global-domain results

• Interesting to compare global time-slice

results to nested model results

• Nice to have global-domain results

• Interesting to compare global time-slice

results to nested model results

Advantages/disadvantages Advantages/disadvantages vs.vs. nested model approachnested model approach

Advantages/disadvantages Advantages/disadvantages vs.vs. nested model approachnested model approach

Advantages:• Nice to have global-domain results.• Needed input data (SST + sea ice extents) are minimal, and universally available. • Results are not subject to degradation by biases in lateral boundary conditions.

Disadvantages:• Regional-scale results are not constrained by lateral boundary conditions.• More demanding of CPU.• Larger volume of output data.

Advantages:• Nice to have global-domain results.• Needed input data (SST + sea ice extents) are minimal, and universally available. • Results are not subject to degradation by biases in lateral boundary conditions.

Disadvantages:• Regional-scale results are not constrained by lateral boundary conditions.• More demanding of CPU.• Larger volume of output data.

What model did I use?What model did I use?What model did I use?What model did I use?

• Fine-resolution version of NCAR CAM3.1 global atmospheric model

• Finite Volume dynamical core• 0.625 deg. (longitude) x 0.5 deg. (latitude)

grid spacing• Ad hoc retuning of parameterizations

performed in collaboration with Hack et al. of NCAR

• Fine-resolution version of NCAR CAM3.1 global atmospheric model

• Finite Volume dynamical core• 0.625 deg. (longitude) x 0.5 deg. (latitude)

grid spacing• Ad hoc retuning of parameterizations

performed in collaboration with Hack et al. of NCAR

1.1. ““Control” or “AMIP” simulationControl” or “AMIP” simulation1.1. Covers 1979-2000Covers 1979-20002.2. driven by observed SSTs and sea ice extentsdriven by observed SSTs and sea ice extents

2.2. ““Future” or “A2” simulationFuture” or “A2” simulation1.1. Covers 2041-2060Covers 2041-20602.2. Driven byDriven by

SST = SSTSST = SSTobsobs + SST + SSTccsmccsmfuturefuture - SST - SSTccsmccsm

presentpresent..

SSTSSTccsmccsm from simulation of A2 emissions scenario from simulation of A2 emissions scenario performed with coarse-resolution version of CCSMperformed with coarse-resolution version of CCSM

3. This method of deriving SSTs provides first-order 3. This method of deriving SSTs provides first-order correction of biases in SSTs of CCSM modelcorrection of biases in SSTs of CCSM model

1.1. ““Control” or “AMIP” simulationControl” or “AMIP” simulation1.1. Covers 1979-2000Covers 1979-20002.2. driven by observed SSTs and sea ice extentsdriven by observed SSTs and sea ice extents

2.2. ““Future” or “A2” simulationFuture” or “A2” simulation1.1. Covers 2041-2060Covers 2041-20602.2. Driven byDriven by

SST = SSTSST = SSTobsobs + SST + SSTccsmccsmfuturefuture - SST - SSTccsmccsm

presentpresent..

SSTSSTccsmccsm from simulation of A2 emissions scenario from simulation of A2 emissions scenario performed with coarse-resolution version of CCSMperformed with coarse-resolution version of CCSM

3. This method of deriving SSTs provides first-order 3. This method of deriving SSTs provides first-order correction of biases in SSTs of CCSM modelcorrection of biases in SSTs of CCSM model

I performed two simulationsI performed two simulationsI performed two simulationsI performed two simulations

1.1. All quantities specified in NARCCAP All quantities specified in NARCCAP protocolprotocol

2.2. Additional monthly-mean stuffAdditional monthly-mean stuff3.3. 3-hourly 3-d atmospheric fields needed to 3-hourly 3-d atmospheric fields needed to

drive a nested atmospheric model. (This is drive a nested atmospheric model. (This is 80% of the data volume).80% of the data volume).

Raw data volume: 40 TbyteRaw data volume: 40 Tbyte After interpolation to specified pressure After interpolation to specified pressure

levels: 65 Tbyte.levels: 65 Tbyte.

1.1. All quantities specified in NARCCAP All quantities specified in NARCCAP protocolprotocol

2.2. Additional monthly-mean stuffAdditional monthly-mean stuff3.3. 3-hourly 3-d atmospheric fields needed to 3-hourly 3-d atmospheric fields needed to

drive a nested atmospheric model. (This is drive a nested atmospheric model. (This is 80% of the data volume).80% of the data volume).

Raw data volume: 40 TbyteRaw data volume: 40 Tbyte After interpolation to specified pressure After interpolation to specified pressure

levels: 65 Tbyte.levels: 65 Tbyte.

What output did I save?What output did I save?What output did I save?What output did I save?

Nested regional model at 9 km driven by global model at ~100 km

cm/yr

“Observations”

1.1. Simulations are complete.Simulations are complete.

2.2. Interpolation to specified atmospheric pressure Interpolation to specified atmospheric pressure levels is complete.levels is complete.

3.3. Conversion to CF-compliant format is not Conversion to CF-compliant format is not complete (although results are already in netcdf complete (although results are already in netcdf format)format)

4.4. AMIP results reside in NERSC archival storageAMIP results reside in NERSC archival storage

5.5. A2 results reside in NCAR mass storageA2 results reside in NCAR mass storage

6.6. It’s difficult to do anything with this much data!It’s difficult to do anything with this much data!

1.1. Simulations are complete.Simulations are complete.

2.2. Interpolation to specified atmospheric pressure Interpolation to specified atmospheric pressure levels is complete.levels is complete.

3.3. Conversion to CF-compliant format is not Conversion to CF-compliant format is not complete (although results are already in netcdf complete (although results are already in netcdf format)format)

4.4. AMIP results reside in NERSC archival storageAMIP results reside in NERSC archival storage

5.5. A2 results reside in NCAR mass storageA2 results reside in NCAR mass storage

6.6. It’s difficult to do anything with this much data!It’s difficult to do anything with this much data!

Status of simulations, etc.Status of simulations, etc.Status of simulations, etc.Status of simulations, etc.

AMIP simulation resembles AMIP simulation resembles planet earthplanet earth

AMIP simulation resembles AMIP simulation resembles planet earthplanet earth

Reference height temperature over landReference height temperature over land

AMIP Precipitation…AMIP Precipitation…AMIP Precipitation…AMIP Precipitation…

CAM CAM vsvs: GPCP : GPCP Legates & WilmottLegates & Wilmott Xie/Arkin Xie/Arkin

AMIP annual reference height AMIP annual reference height temperaturetemperature

AMIP annual reference height AMIP annual reference height temperaturetemperature

CAM CAM vs.vs. Legates Legates & Wilmott (1920-& Wilmott (1920-1980)1980)

AMIP annual reference height AMIP annual reference height temperaturetemperature

AMIP annual reference height AMIP annual reference height temperaturetemperature

CAM CAM vs.vs. Wilmott Wilmott & Matsura & Matsura (1950-1999)(1950-1999)

DJFJJA

JJA

Reference height temperature biasesReference height temperature biasesReference height temperature biasesReference height temperature biases

DJF JJA

JJADJF

DJF JJAJJA

NCAR 2.5 x 2.0 LLNL 0.625 x 0.5

Biases in JJA temperatures are inherited from Biases in JJA temperatures are inherited from NCAR coarse-resolution model versionNCAR coarse-resolution model version

Biases in JJA temperatures are inherited from Biases in JJA temperatures are inherited from NCAR coarse-resolution model versionNCAR coarse-resolution model version

Errors in JJA TREFHTErrors in JJA short-wave cloud forcing

Temperature biases seem to Temperature biases seem to result from cloud errorsresult from cloud errors

Temperature biases seem to Temperature biases seem to result from cloud errorsresult from cloud errors

Anomalies in daily maximum near-surface temperaturesAnomalies in daily maximum near-surface temperaturesAnomalies in daily maximum near-surface temperaturesAnomalies in daily maximum near-surface temperatures

Anomalies in daily minimum near-surface temperaturesAnomalies in daily minimum near-surface temperaturesAnomalies in daily minimum near-surface temperaturesAnomalies in daily minimum near-surface temperatures

DJF JJAJJA

AMIP seasonal precipitation biasesAMIP seasonal precipitation biasesAMIP seasonal precipitation biasesAMIP seasonal precipitation biases

DJF JJA

Volumetric precipitation distribution on native grids

0

1

2

3

4

5

6

7

8

1 5 10 20 40 80 80+

Precip. class (mm/day)

(km^3/day)

NOAA observations

GPCP observations

2.0 x 2.5 deg.

2.0 x 2.5 deg 64-cellCSRM

2.0 x 2.5 deg 32-cellCSRM

2.0 x 2.5 deg. 128-cell CSRM

1.0 x 1.25 deg.

1.0 x 1.25 deg.CSRM

0.5 x 0.625 deg.

o

Daily precipitation Daily precipitation amountsamounts

Daily precipitation Daily precipitation amountsamounts

Observations

QuickTime™ and aCinepak decompressor

are needed to see this picture.

Summary: LLNL time-slice simulationsSummary: LLNL time-slice simulationsSummary: LLNL time-slice simulationsSummary: LLNL time-slice simulations

• Next time I’ll be smarter about the Next time I’ll be smarter about the difficulties of handling 60+ Tbyte of difficulties of handling 60+ Tbyte of output.output.

• Results look like planet earth, but...Results look like planet earth, but...

• … …Near-surface temperatures have large Near-surface temperatures have large biases in some regions, especially in biases in some regions, especially in summer.summer.

• These seem to be related to cloud These seem to be related to cloud errors and are inherited from the coarse-errors and are inherited from the coarse-resolution model version. resolution model version.

• Daily temperatures and precipitation Daily temperatures and precipitation amounts are simulated better than in amounts are simulated better than in coarser-resolution versions of the same coarser-resolution versions of the same model.model.

AMIP annual precipitable waterAMIP annual precipitable waterAMIP annual precipitable waterAMIP annual precipitable water

CAM minus MODIS

CAM minus

ECMWF

CAM minus CAM minus NCEPNCEP

50 km

QuickTime™ and aCinepak decompressor

are needed to see this picture.

300 km

QuickTime™ and aCinepak decompressor

are needed to see this picture.