Late 20th Century (and Future) Simulations of Western Atlantic Winter Storms in Several CMIP5 Models and other Regional Climate Ensembles Brian A. Colle,

Late 20th Century (and Future) Simulations of Western Atlantic Winter Storms in Several CMIP5 Models and other Regional Climate Ensembles

Brian A. Colle, Zhenhai Zhang, Ping Liu, Kelly Lombardo, Edmund Chang, and Minghua Zhang,

Stony Brook University - SUNY

MotivationDetermine how well the CMIP5 models can simulate the

Western Atlantic extratropical cyclones (density, intensity, genesis, deepening, etc…) for the cool season (Nov-March) for 1979-2004 historical period.

What is the impact of model resolution and weighting the CMIP ensemble members based on past performance?

Is there any indication of future cyclone change? Frequency, intensity, or spatial distribution?

Data and Method

Data - Mean Sea Level Pressure (MSLP)

CFSR: Climate Forecast System Reanalysis from NCEP, 1979~2004, 6-houly (Also tried ERA-Interim)

CMIP5: Coupled Model Intercomparison Project Phase 5, new set of climate model experiments for IPCC AR5.

Historical - 1979~2004, 6-hourly

Rcp8.5 - high emissions scenario, 2006~2098, 6-hourly

Models Total Grids LAT Grids LON GridsCESM 55296 192 288

EC-EARTH 51200 160 320MRI-CGCM3 51200 160 320CNRM-CM5 32768 128 256

MIROC5 32768 128 256HadGEM2-ES 27840 145 192HadGEM2-CC 27840 145 192

INMCM4 21600 120 180IPSL-CM5A-MR 20592 143 144

MPI-ESM-LR 18432 96 192NorESM1 13824 96 144

GFDL-ESM2M 12960 90 144IPSL-CM5A-LR 9216 96 96

BCC-CSM1 8192 64 128MIROC-ESM-C 8192 64 128

High Resolution: 7 ModelsLow Resolution: 8 Models

Total Grids = LatGrids × LonGrids

CMIP5 _ 15 models

Methods

Tracker Procedure - Hodges TRACK Scheme

(A General Method For Tracking Analysis And Its Application To Meteorological Data; Hodges, 1994)

Step I: Spectral Bandpass FilterRemove planetary scale background and truncate too small scales, and get

MSLP anomaly field.

Step II: Cyclone Tracking Separate the data points into object and background points, locate the

minimum pressure; Use a constrained optimization of a cost function to determine the next potential cyclone center position; Label these connected cyclone centers with a unique storm ID.

Main Parameters

Wavelength (600 – 10000km)

removing the planetary scales and too small scales

Minimum Lifetime (24 hours)Minimum moving distance (1000km)

filtering centers exist for too short time or remain too stationary

Also compared with the Hodges 850 hPa Vorticity Tracking Approach Too many non-cyclones.

Cyclone Center

Total Centers

Centers Missed

False Alarm

numbers 2286 181 103

rate / 7.9% 4.5%

Verification – 11 Januariesof CFSR from 1980-2000 (every other

January)Lon 90-40°W, Lat 20-60°N

Some ResultsCool SeasonNovember – March, 5 months

Historical part1979 – 2004 (26 cool seasons)

3 Future partsEarly 21st Century 2009-2038Middle 21st Century 2039-2068Late 21st Century 2069-2098

Large blue box: for spacial density plotSmall dashed box: for statistical analysis

CycloneTrack DensityCFSR_1979-2004

7 Higher Resolution Members

Cyclone numbersper cool season

per 2.5° x 2.5°

CMIP5_Mean/Spread1979-2004

8 Lower Resolution Members

Cyclone Intensity Distribution 8 Lower Resolution Members

Cyclone Intensity Distribution 7 Higher Resolution Members

CMIP5 Cyclone RankingRank (red – high resolution)

ModelsCombined

RankTrack Density

CorrelationIntensity

Distribution Resolution

EC-EARTH 2.5 1 4 160x320

MPI-ESM-LR 4.0 2 6 96x192

HadGEM2-ES 4.5 4 5 145x192

MRI-CGCM3 5.0 9 1 160x320

CESM 5.5 8 3 192x288

CNRM-CM5 5.5 3 8 128x256

HadGEM2-CC 6.0 10 2 145x192

NorESM1 8.0 7 9 96x144

BCC-CSM1 9.0 11 7 64x128

GFDL-ESM2M 9.0 6 12 90x144

IPSL-CM5A-MR 9.5 5 13 143x144

MIROC-ESM-CHEM 12.0 14 10 64x128

INMCM4 13.0 15 11 120x180

MIROC5 13.0 12 14 128x256

IPSL-CM5A-LR 14.0 13 15 96x96

Future Cyclone Density Change “Best 8”

GenesisDensityCyclone numbersper 5 cool seasons

per 2.5° x 2.5°

Best 8 members

CFSR_1979-2004 CMIP5_1979-2004

CMIP5_2009-2038 CMIP5_2039-2068 CMIP5_2069-2098

Cyclone Density from

50-km NARCCAP

North American Regional Climate Change

Assessment Program (DJF

1979-1999) forced by NCEP2 analysis

Challenges: Sensitivity of Regional Climate Models to Physics (20-km WRF forced by NCEP2 reanalysis)

Number of DJF cyclones per 2.5x2.5deg (1985-2004) CFSR WRF-mean WRF-KF/YSU

WRF-BM/YSUWRF-KF/MYJWRF-BM/MYJ WRF-GR/YSU

WRF-GR/QNSE

SummaryThe 15 CMIP5 models can realistically predict the 1979-

2004 winter cyclone density distribution over the western Atlantic and U.S. East Coast; however, most of the CMIP5 models underpredict the magnitude.

The “higher resolution” CMIP5 models better simulate the cyclone density and intensity, but most models underpredict the relatively deep cyclones (< 980 hPa).

The cyclone numbers decrease gradually over western Atlantic and U.S. East Coast in the future period (2009-2098) for Rcp8.5 experiment. There is a reduction in cyclone genesis along East Coast.

A diverse set of physics (or multi-models) is required for regional ensembles (plus ensemble weighting) given the 20-30% uncertainties in predicted cyclone climatology related to model physics uncertainties.