narccap multi-model simulations: initial ncep-driven results

NARCCAP Users MeetingNARCCAP Users Meeting 14-15 February 200814-15 February 2008

NARCCAP Multi-Model Simulations:NARCCAP Multi-Model Simulations:Initial NCEP-Driven ResultsInitial NCEP-Driven Results

Bill Gutowski, Bill Gutowski, Ray ArrittRay Arritt, Gene Takle, , Gene Takle, Dave FloryDave Flory, , John BaranickJohn Baranick

(Iowa State University)(Iowa State University)andand

The NARCCAP Modelers TeamThe NARCCAP Modelers Team


NARCCAP ParticipantsNARCCAP Participants

• Raymond Arritt, David Flory, William Gutowski, Gene Takle, Iowa State University, USA

• Richard Jones, W. Moufouma-Okia, Hadley Centre, UK• Daniel Caya, Sébastien Biner, OURANOS, Canada• David Bader, Phil Duffy, Lawrence Livermore National Laboratories, USA• Filippo Giorgi, Abdus Salam ICTP, Italy• Isaac Held, NOAA Geophysical Fluid Dynamics Laboratory, USA• René Laprise, Univ. de Québec à Montréal, Canada• Ruby Leung, Y. Qian, Pacific Northwest National Laboratories, USA• Linda Mearns, Don Middleton, Doug Nychka National Center for Atmospheric

Research, USA• Ana Nunes, John Roads, Scripps Institution of Oceanography, USA• Steve Sain, Univ. of Colorado at Denver, USA• Lisa Sloan, Mark Snyder, Univ. of California at Santa Cruz, USA


Comparison with observationsComparison with observations

• ObservationsObservations Precip: University of Washington VIC retrospective analysisPrecip: University of Washington VIC retrospective analysis

500 hPa Heights: North American Regional Reanalysis500 hPa Heights: North American Regional Reanalysis

• Comparison period: 1980-1999Comparison period: 1980-1999 1979 omitted - (a) spinup (b) WRFP began 1 Sep 791979 omitted - (a) spinup (b) WRFP began 1 Sep 79 UW data end in mid-2000UW data end in mid-2000

• Analyses: monthly mean precipitation & 500 hPa ZAnalyses: monthly mean precipitation & 500 hPa Z Fields received at Iowa State for format checkFields received at Iowa State for format check

For several regions in the U.S. (UW analysis extends to ~ 53For several regions in the U.S. (UW analysis extends to ~ 53ooN)N)


Regions AnalyzedRegions Analyzed


Coastal CaliforniaCoastal California

• Mediterranean climate: wet winters and dry summers (Koeppen types Csa, Csb)

• ENSO can have strong effects on interannual variability of precip


Monthly precipitation for coastal California

0

2

4

6

8

10

12

14

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3 MM5I

ECPC MRCC

Observed (UW) Ensemble Mean

1982-83 El Nino

1997-98 El Nino

multi-year drought

Monthly Time Series - Coastal CAMonthly Time Series - Coastal CA


Time Series Correlations - Coastal CATime Series Correlations - Coastal CA

Model Correlation

RCM3 0.946

MM5I 0.946

ECPC 0.966

MRCC 0.959

Ensemble 0.968


0

2

4

6

8

10

12

14

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3 MM5I

ECPC MRCC



Mean annual cycle, coastal California

0

1

2

3

4

5

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

mm/day

RCM3 MM5I

ECPC MRCC

Obs (UW) Ensemble

Monthly mean precipitation, mm/day

0

2

4

6

8

10

12

0 2 4 6 8 10 12

Observed

Ensemble mean

Further Statistics - Coastal CAFurther Statistics - Coastal CA


1997-1998 El Ni1997-1998 El Niññoo

• Strongest El Niño in the instrumental record.


October 1997RegCM3

MRCC MM5

RSM

Observed (CRU)


January 1998RegCM3

MRCC MM5

RSM

Observed (CRU)


March 1998RegCM3

MRCC MM5

RSM

Observed (CRU)



0

2

4

6

8

10

12

14

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3 MM5I

ECPC MRCC


Circulation with Extreme PrecipitationCirculation with Extreme PrecipitationPrecip. Max.: Feb 1998 Precip. Max.: Feb 1998


Simulated Simulated Circulation Circulation

with Extreme with Extreme PrecipitationPrecipitation


Deep SouthDeep South

• Humid mid-latitude climate with little seasonality in precip amount (Koeppen type Cfa).

• Past studies have found problems with RCM simulations of cool-season precip in this region.


Monthly precipitation for Deep South

0

2

4

6

8

10

12

14

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3 MM5IECPC MRCCObserved (UW) Ensemble Mean

Monthly Time Series - Deep SouthMonthly Time Series - Deep South



0

2

4

6

8

10

12

14

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3MM5IECPCMRCCObserved (UW)Ensemble Mean

ModelModel CorrelationCorrelation

RCM3 0.257

MM5I 0.377

ECPC 0.636

MRCC 0.645

Ensemble 0.597

Ensemble of MRCC and ECPC

0.709

ECPC and MRCC both incorporate ECPC and MRCC both incorporate large-scale information in the domain large-scale information in the domain interior: ECPC is a perturbation model interior: ECPC is a perturbation model (RSM), while MRCC uses spectral (RSM), while MRCC uses spectral nudging. nudging.

Time Series Correlations - Deep SouthTime Series Correlations - Deep South


Mean annual cycle, Deep South

0

1

2

3

4

5

6


mm/dayRCM3 MM5I

ECPC MRCC

Obs (UW) Ensemble


0

2

4

6

8

10

0 2 4 6 8 10Observed

Ensemble mean

too dry in the cool half of the year

high rates (> 4 mm/day) are underpredicted

Further Statistics - Deep SouthFurther Statistics - Deep South



0

2

4

6

8

10

12

14

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3MM5IECPCMRCCObserved (UW)Ensemble Mean

Circulation with Extreme PrecipitationCirculation with Extreme PrecipitationPrecip. Max.: April 1991 Precip. Max.: April 1991


Simulated Simulated Circulation Circulation

with Extreme with Extreme PrecipitationPrecipitation


Comments and speculationsComments and speculations

• A simple unweighted ensemble mean usually A simple unweighted ensemble mean usually performs better than the best individual model, or performs better than the best individual model, or close to the best model when spread is large.close to the best model when spread is large.

• Hypothesis: Downscaling of ENSO could be an Hypothesis: Downscaling of ENSO could be an especially suitable use for a coupled GCM-RCM:especially suitable use for a coupled GCM-RCM:• RCMs perform well in coastal California during ENSORCMs perform well in coastal California during ENSO• Some AOGCMs can produce reasonable ENSO (see e.g., Some AOGCMs can produce reasonable ENSO (see e.g.,

Van Oldenborgh et al. 2005).Van Oldenborgh et al. 2005).• Two of these AOGCMs are used in NARCCAP: GFDL Two of these AOGCMs are used in NARCCAP: GFDL

CM2.1 and HadCM3.CM2.1 and HadCM3.


Comments and speculationsComments and speculations

• Incorporation of large-scale information into the RCM Incorporation of large-scale information into the RCM (whether through spectral nudging or use of a (whether through spectral nudging or use of a perturbation form of the governing equations) perturbation form of the governing equations) appears to be an advantage for the Deep South appears to be an advantage for the Deep South region.region.• This advantage does not carry over to other regions (or This advantage does not carry over to other regions (or

is outweighed by other factors, e.g., model physics).is outweighed by other factors, e.g., model physics).

5-Yr Return Period Amounts5-Yr Return Period Amounts[mm/day][mm/day]

OBSERVATIONSOBSERVATIONS

1980-19991980-1999

Thanks to Hayley Fowler for diagnostic codeThanks to Hayley Fowler for diagnostic code

5-Yr Return 5-Yr Return Period Period

AmountsAmounts[mm/day][mm/day] OBSOBS

MM5-PNLMM5-PNL

MM5-ISUMM5-ISU

MRCCMRCC

ECPCECPC

RegCM3RegCM3


Thank YouThank You

Further InformationFurther Information

1. General: 1. General: http://www.narccap.ucar.eduhttp://www.narccap.ucar.edu

2. Archive Information: 2. Archive Information: http://rcmlab.agron.iastate.edu/narccap/output_archive.htmlhttp://rcmlab.agron.iastate.edu/narccap/output_archive.html

3. Data portal: http://esg.ucar.edu/forward.htm?forward=narccap


EXTRAS


U.S. Corn Belt

• Western part is a continental climate with warm summers and cold winters becoming less continental to the east. (Koeppen types Dfa, Dfb)

• Maximum precipitation in April-June• seasonality of precip is important for agriculture, e.g.,

drawdown of soil moisture during the growing season

• Includes the Upper Mississippi River basin


0

1

2

3

4

5

6

7

8

9

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

Ensemble

Observed (UW) 1993 flood1988 drought

Model Correlation

RCM3 0.722

MM5I 0.692

ECPC 0.636

MRCC 0.760

Ensemble 0.788

Monthly precipitation, U.S. Corn Belt region

0

1

2

3

4

5

6

7

8

9

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3 MM5I

ECPC MRCC

Ensemble Observed (UW)


too wet in winter/spring

too dry in summer

low rates (< 2 mm/day) are overpredicted

high rates (> 4 mm/day) are underpredicted


Upper Mississippi River Basin

• Continental climate with warm summers and cold winters (Koeppen types Dfa, Dfb).

• Maximum precipitation in April-June:• seasonality of precip is important for impacts, e.g.,

drawdown of soil moisture during the growing season

• Most NARCCAP models simulated this region in the PIRCS project.


Monthly precipitation, Upper Mississippi River basin

0

1

2

3

4

5

6

7

8

9

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

RCM3 MM5I

ECPC MRCC

Ensemble Observed (UW)

0

1

2

3

4

5

6

7

8

9

1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

mm/day

Ensemble

Observed (UW)

1993 flood1988 drought

Model Correlation

RCM3 0.745

MM5I 0.696

ECPC 0.627

MRCC 0.779

Ensemble 0.791


Mean annual cycle, Upper Mississippi River basin

0

1

2

3

4

5


mm/day

RCM3 MM5I

ECPC MRCC

Obs (UW) Ensemble

too wet in winter/spring

too dry in summer


0

2

4

6

8

0 2 4 6 8

Observed

Ensemble mean

low rates (< 2 mm/day) are overpredicted

high rates are underpredicted


1982-1983 and 1997-98 El Niños

• Some AOGCMs give realistic simulation of ENSO:• see e.g., van Oldenborgh et al. (2005, Ocean Science) for

IPCC AR4 models

• Do regional models give realistic precipitation during El Niño events?• If so, perhaps a combined AOGCM-RCM approach can

give useful results for ENSO in future climates.

• We examine evolution of precipitation during the 1982-1983 and 1997-1998 El Niño: onset, peak, and withdrawal.


October 1982RegCM3

MRCC MM5

RSM

Observed (CRU)


February 1983RegCM3

MRCC MM5

RSM

Observed (CRU)


April 1983RegCM3

MRCCMM5

RSM

Observed (CRU)


1993 flood in the north-central U.S.1993 flood in the north-central U.S.

• The event mainly was the result of synoptic-mesoscale dynamics with little direct influence by terrain.

• This event was the subject of an early regional model intercomparison which performed 60-day simulations (PIRCS, 1B).


Summer 1993 flood in central U.S.RegCM3

MRCC MM5

RSM

Observed (CRU)

July 1993 precipitation

• model skill is very similar to PIRCS 1B


AmountsAmounts[mm/day][mm/day]

OBSOBS

ECPCECPC RegCM3RegCM3



OBSOBS

MM5-PNLMM5-PNL MRCCMRCC



OBSOBS

MM5-PNLMM5-PNL MM5-ISUMM5-ISU

narccap multi-model simulations: initial ncep-driven results

Documents

extrasnarccap users

narccapnarccap users

continental climate

archive information

growing seasonincludes

hayley fowler

cold winters

dfbmaximum precipitation