RCM Climate Change Projection Uncertainty

Xin-Zhong LiangDepartment of Atmosphere & Ocean ScienceEarth System Science Interdisciplinary CenterUniversity of Maryland, College Park

2013 August 13

NCPP Quantitative Evaluation of Downscaling Workshop

Only MRI with a medium resolution captures the observed NAM rainfall annual cycle. CCSM3 with higher resolution and updated physics reduces PCM amplitude error by half. MIROC3 finer resolution improves the phase but worsens the magnitude .

Liang, X.-Z., J. Zhu, K.E. Kunkel, M. Ting, and J.X.L. Wang, 2008: Do CGCMs simulate the North American monsoon precipitation seasonal-interannual variability? J. Climate, 21, 4424-4448 .

Resolution Increase Does Not Solve All Problems

CWRF Terrestrial HydrologyKanawha

Kentucky

Green

Tennessee

Kanawha

Kentucky

Green

Tennessee

0

5

10

15

20

25

30

0 30 60 90 120 150 180 210 240 270 300 330 360Days Since Jan 1995

Stre

am F

low

(mm

/day

) 0

50

100

150

200

Pre

cipi

tatio

n (m

m/d

ay)Precipitation

CLMCLM+CSSObserved

USGS Sta. 03320000Green River

Choi 2006; Choi et al. 2007; Choi and Liang 2010; Yuan and Liang 2010; Liang et al. 2010; Choi et al. 2013

D1

D2

D3

Urban and Built-upDryland Crpland and PastureIrrigated Cropland and PastureCropland/Grassland MosaicCropland/Woodland MosaicGrasslandShrublandMixed Shrubland/GrasslandSavanna

Deciduous Broadleaf ForestEvergreen Broadleaf ForestEvergreen Needleleaf ForestMixed ForestWater BodiesWooded WetlandBarren or Sparsely VegetatedWooded TundraMixed Tundra

atmosphere

land

ICs

LBCs

SST

CFS

NARR

ocea

n

CWRF Downscaling Seasonal Climate Prediction over the U.S.

CWRFMLO

CAMGFS

CFSSST ICs

NOAA2008-2011

Monthly

Daily

Anderson, B.T., K. Hayhoe, and X.-Z. Liang, 2009: Anthropogenic-induced changes in the 21st Century summertime hydroclimatology of the Northeastern US., Climate Change, doi.10.1007/s10584-009-9673-3.

RCM Better Resolves Extremes

Northeast U.S. Assessment

Number of Rainy Days Rain Intensity Daily Rainfall 95th Percentile 1993 JJA Mean

OBS

ERI

CWRF

The reanalysis has already assimilated local observational data, while CWRF is driven by only LBCs. The CWRF skill will be enhanced if assimilating local data.

CWRF with ECP/W closure over the U.S. land

a) Spatial frequency distributions of root mean square errors (RMSE, mm/day) predicted by the CFS and downscaled by the CWRF and b) CWRF minus CFS differences in the equitable threat score (ETS) for seasonal mean precipitation interannual variations. The statistics are based on all land grids over the entire inner domain for DJF, JFM, FMA, and DJFMA from the 5 realizations during 1982-2008. From Yuan and Liang 2011 (GRL).

CWRF Improves Seasonal Climate Prediction

Statistic Metric for RCM Added Value

Consider differences in statistical distribution of skill, not just changes of mean skill.

Construct PDFs of skill over a region for both global and regional models. Focus on the part of the distribution above a critical skill level (Sc) representing minimum useful skill.

Sc

Global model (solid)

Regional model (dashed)

Added Value Index (AVI) is the area between the curves beyond the crossover point where regional model skill begins to exceed global model skill (if the crossover point exists).

Interpretation: Regional model skill is higher than global model over AVI % of the domain for the corresponding skill range.

Crossover point

After Kanamitsu and DeHaan, 2011, J. Geophys. Res., doi:10.1029/2011JD015597

Curtsey of Raymond Arritt

Interannual CORR over USA

Why Do RCM Results Differ?

• Domain: U.S. + Adjacent for CWRF & CMM5,Extended North America for NARCCAP

• Resolution: 30 km for CWRF & CMM5,50 km for all other

NARCCAP RCMs• Forcing: linear-exp relaxation in buffer zones of

14 (CWRF, CMM5), 10 (WRFG) gridslinear relaxation in 4 grids (MM5I, HRM3)domain spectral nudging (ECP2, CRCM)NARCCAP IA correlations differ largelydue to the strength of forcing integrated

• Physics: CWRF is much better than CMM5,being identical in all other settings

Different dynamics may also contribute

Climate Change Uncertainty• Biases from incomplete physics in GCMs & RCMs

• Climate sensitivities of GCMs & RCMs

• Projecting emissions scenarios

• Natural variability

Climate Change Projection by GCMs and RCMs

Propagation of GCM Present Climate Biases into Future Change Projections: Temperature

Liang, X.-Z., K.E. Kunkel, G.A. Meehl, R.G. Jones, and J.X.L. Wang, 2008: RCM downscaling analysis of GCM present climate biases propagation into future change projections. Geophys. Res. Lett., 35, L08709, doi:10.1029/2007GL032849.

U.S. + Mexico-7-6-5-4-3-2-1012345

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

Precipitation Temperature

CU BS AS

Central Great Plains-7-6-5-4-3-2-1012345

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

Precipitation Temperature

Central U.S.-7-6-5-4-3-2-1012345

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

Precipitation Temperature

Southeast U.S.-5-4-3-2-101234567

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

PGR-PCM

PKF-PCM

HGR-HAD

HKF-HAD

Precipitation Temperature

Projected U.S. Heat Wave Changes

-10

0

10

20

30

-10

0

10

20

30

-4

0

4

8

-4

0

4

8 T

empe

ratu

re In

crea

se (°

C)

Tem

pera

ture

Incr

ease

(°C

)

Diff

eren

ce in

Ann

ual N

umbe

r

D

iffer

ence

in A

nnua

l N

umbe

r

c) Chicago

d) Northeast US

10%

1% 1%

10%

10%

1%

a) Chicago

1%

1%

1%

10%

b) Northeast US

1%1%

1%

10%

-4

0

4

8

12

RCM-P RCM-H OBSScenario, Period

0

2

4

6

Scenario, Period

Kunkel, K.E., X.-Z. Liang, and J. Zhu, 2010: Regional climate model projections and uncertainties of U.S. summer heat waves. J. Climate, 23, 4447-4458.

Projections of changes in the average annual 3-day heat wave temperature (°C) for a) Chicago and b) Northeast US and of the annual average number of heat wave days for c) Chicago and d) Northeast US. The two sets of bars on the far left side of a) and b) compare the present-day annual 3-day heat wave temperature spread (from its own summer mean temperature as simulated and observed); and model biases (from observations). The simulations are arranged from left to right in order of increasing greenhouse gas concentrations. The % number at the bar top depicts the corresponding statistical significance level.

MKF MGR

EOPOBS

Optimized Physics-Ensemble Prediction

Liang et al. (2007, JCL)

Optimized Physics Ensemble Prediction of Precipitation

In summer 1993

Spatial frequency distributions of correlations (top) and rms errors (bottom) between CWRF and observed daily mean rainfall variations in

summer 1993. Each line depicts a specific configuration in group of the five key

physical processes (color). The ensemble result (ENS) is the average of all runs with

equal (Ave) or optimal (OPT) weights, shown as black solid or dashed line.

The physics ensemble mean substantially increases the skill score over individual configurations, and there

exists a large room to further enhance that skill

through intelligent optimization.

Liang et al. (2012, BAMS)

What Should We Do?• Structure Uncertainty:

- Multiple GCMs with (low, medium, high) climate sensitivities representative of the possible range

- Multiple RCMs or physics configurations with notable added values key to the application

- Multi-model ensemble with optimal weighting based on present-day climate biases

• Emission Scenario:- IPCC Representative Concentration Pathways:

RCP 8.5 & 4.5, or also RCP 6.0 & 2.6 if affordable- For projection to 2050, RCP 8.5 & 4.5 is desirable