The ECHAM-SIT-TIMCOM Model (EHTW ESM) and

potential applications

Yangtze

River

Taiwan West Pacific Global Forecast System Planning Workshop

臺灣與西北太平洋氣候預測全球模式發展規劃研討會中央氣象局310會議室

CWB, Taipei, Taiwan, May 8-9, 2013

Ben-Jei Tsuang (莊秉潔), Department of Environmental Engineering, NCHU, Taiwan

Yu-Heng Tseng (曾于恒), NCAR, USA, NTU, Taiwan

Question? Can a coupled OAGCM model do a

better weather and climate forecast than an AGCM

driven by a good SST?

AGCM: Two-Tier Approach

1. SST prediction by statistical/physical model

2. SST used as the boundary condition to drive AGCM

OAGCM: One-Tier Approach

1. Prescribed ICs for both atmosphere and ocean

2. Run the coupled model with/without flux correction

(nudging)

Answer? (No/Yes) No, if the bias of the mean SST is too large by OAGCM, the

resolved shorter-time scale flux is not good enough.

1. ENSO can be well predicted by the two-tier approach.

2. Single coupled OAGCM usually can not beat multiple methods for

SST prediction in ENSO time-scale.

Yes, only the bias from the SST mean state by the coupled model

can be overcomed by a better resolving the air-sea interaction with

time scale short than the mean state.

1. For example, if the rainfall by diurnal cycle and Mudian-Jullian

Oscillation (MJO) can be better resolved by the coupled model.

2. 𝐿𝐸 = 𝑤 ∙ 𝑞 = 𝑤 ∙ 𝑞 + 𝑤′ ∙ 𝑞′𝑚𝑜

+ 𝑤′ ∙ 𝑞′𝑚𝑗𝑜

+ 𝑤′ ∙ 𝑞′ℎ

Test the idea!

AGCM run (prescribed SST, AMIP-type run) (ECHAM)

OAGCM run (ECHAM/SIT/TIMCOM)

AGCM + 1 d ocean model with nudging WT (ECHAM/SIT)

5

AGCM: ECHAM5 The dynamical part of ECHAM is formulated in spherical harmonics. After the inter-

model comparisons by Jarraud et al. (1981) and Girard and Jarraud (1982) truncated expansions in terms of spherical harmonics were adopted for the representation of dynamical fields.

The transform technique developed by Eliasen et al. (1970), Orszag (1970), and Machenhauer and Rasmussen (1972) is used such that non-linear terms, including parameterizations, are evaluated at a set of almost regularly distributed grid points - the Gaussian grid.

In the vertical, a flexible coordinate is used, enabling the model to use either the usual terrain following sigma coordinate (Phillips, 1957), or a hybrid coordinate for which upper-level model surfaces atten over steep terrain, becoming surfaces of constant pressure in the stratosphere (Simmons and Burridge (1981) and Simmons and Struring (1981)).

Moist processes are treated in a different way using a mass conserving algorithm for the transport (Lin and Rood, 1996) of the different water species and potential chemical tracers. The transport is determined on the Gaussian grid.

6

OGCM: TIMCOM

Mixed Arakawa A and C grids; Z-level, rigid-lid/free surface, fourth-order accurate;

Unfiltered etopo2 or etopo5 bathymetry;

< 10,000 lines

History of TIMCOM (3-

D Ocean model)POCM

(1988)

Bryan (1969)

Cox (1970)

Semtner

(1974)

Cox

(1984)

FRAM

(1991)

MOM1

(1990)

MOM2

(1995)

MOM2

(1996)

OCCAM

(1995)

Killworth

et al.(1991)POP

(1992)

POP

(1994)

POP

(1996)

CME

(1989)

CSM

(1996)

NCOM

(1993)

SOMS

(1987)

DIECAST

(1994)

DIECAST

(1997)

TIMCOM

(2010)

CANDIE

(1998)POP2

(2003)

MOM3

(1998)

MOM4

(2003)

LANL UK GFDL NCAR

PGCM

(1991)

TAIWAN

8

SIT (Snow/Ice/Thermocline) Coupler

AtmosphereECHAM (AGCM)19/31 levels(T31-T213)

Ocean:TIMCOM (OGCM)31 Levels (T31-T213)

VD

IFF/SIT (A

ir/Sno

w/Ice/Th

ermo

cline

)2

sno

w+2

ice+4

1 w

ater levels

11 levels in the upper 10

m depth of ocean (0.5

mm, 1, 2, 3, .., 10 m)

Figure 2. ECHAM/SIT/TIMCOM model structure. Note that ocean grid

collocates with the atmospheric grid (or its subdivision equally).

SIT Coupler: based on Governing EquationsContinuity eqn.

Momentum eqn.

Conservation eqn. for temperature and salinity Eqn. of State

Hydrostatic Eqn.

Jan, 1957

Typhoon Morakot(8 August, 2009)

Rainfall, WT and EvapECHAM+3D Ocean (npo0nnnn run) (IC + nudging)

Preliminary Results of present (1945-

2010) climate simulation

AGCM: AMIP run driven by Hadley SST and seaice

(ECHAM)

AGCM+1-D ocean run: prescribed CO2, WT> 10 m depth

nudging with Ishii ocean profile. No nudging within 0-10 m

depth (ECHAM/SIT)

OAGCM run + 10-d nudging: prescribed CO2, WT> 10 m

depth nudging with Ishii ocean profile. No nudging within 0-

10 m depth (ECHAM/SIT/TIMCOM)

OAGCM run + no nudging: prescribed CO2

(ECHAM/SIT/TIMCOM)

PobiNN(ECHAM/SIT/TIMCOM) VS Obstsw

Monthly Mean : Seasonal cycle

Chen, 2013

Chen, 2013

OAGCM+ no

nudging

AGCM+1d ocean Hadley SST

wavelet analysis for NINO3.4

( quick result, to be checked…)

Chen, 2013

OAGCM+ 10-d

nudging

Features: Improved air-sea interaction

for MJO and diurnal time scales

Exchange AGCM and OGCM flux/SST with time scale < 40

min.

11 levels within 10-m depth (for recording the information

of ocean warm layer and cool skin)

AGCM and OGCM grids are co-allocated. (reduced the

interpolation needed)

Gasper et al. vertical diffusivity scheme is used. 1.5

TKE+mixing/dissipation length approach

ECHAM/SIT/TIMCOM (EHTW ESM)

SIT (Snow/Ice/Thermocline) Coupler

AtmosphereECHAM (AGCM)19/31 levels(T31-T213)

Ocean:TIMCOM (OGCM)31 Levels (T31-T213/2 deg)

VD

IFF/SIT (Air/Sn

ow

/Ice/Th

erm

oclin

e)

2 sn

ow

+2 ice+

41

water levels

Figure 9 Simulated x-sliced profiles of

water temperature, salinity, current,

heat diffusivity, turbulent kinetic energy,

mixing length in February 1992 at

51.1875E. (Tsuang et al., 2001)

Heat diffusivity increases with

depth, then drop to background

below mixing depth (Gasper et

al., 1990)

Kh=sqrt(TKE)*mixing length

Gasper et al.’s (1990, JGR) vertical diffusivity

Diurnal SST,

x: month(1~12),

y: hour(0~23 LT)

(a) TAO

(b) SIT_T31

WL Tseng (2012)

Ocean-atmosphere interaction: A key element of the Madden-Julian

Oscillation

Wan-Ling Tseng1, Ben-Jei Tsuang2, Noel Keenlyside3, Huang-Hsiung Hsu4, and Chia-Ying Tu4

1. GEOMAR, Helmholtz Centre for Ocean Research Kiel, Kiel, Germany2. National Chung-Hsing University, Taichung, Taiwan3. University of Bergen, Bergen , Norway 4. Academia Sinica, Research Center for Environmental Changes, Taipei, Taiwan

Final climate simulation test: SST and

precipitation!

AMIP run (1945-2010): Hadley SST and seaice (ECHAM)

AGCM+1-D ocean run (1945-2010): prescribed CO2, WT>

10 m depth nudging with Ishii ocean profile. No nudging

within 0-10 m depth (ECHAM/SIT)

OAGCM run (1945-2010): prescribed CO2, WT> 10 m

depth nudging with Ishii ocean profile. No nudging within 0-

10 m depth (ECHAM/SIT/TIMCOM)

Hadley SST

AGCM+1D ocean

(ECHAM/SIT with

10-d nudging)

OAGCM run

(ECHAM/SIT/TIMCOM

with 10-d nudging)

Mean

Difference

DiffRatio

Std.

SST Simulation vs Hadley SST (T106)

Chen, 2013

GPCP AMIP run

(ECHAM)

OAGCM run

(ECHAM/SIT/TIMCOM

with 10-d nudging)

Mean

Difference

DiffRatio

Std.

Precip. Simulation vs GPCP (T106)

Chen, 2013

T63 AGCM AGCM+1d

ocean

OAGCM with

10-d nudging

OAGCM without

nudging

CORR 0.79 0.72 0.81 0.76

SD_Ratio 1.20 0.95 1.10 1.03

SD_EHTW 2.06 1.63 1.89 1.76

SD_GPCP 1.71 1.71 1.71 1.71

RMS 1.64 1.58 1.26 1.45

EHTW ( T63 ) Global Precip. Simulation : relative to the GPCP climatology for 1979-2009

T63T106

T31

Can a coupled OAGCM model do a better weather and

climate forecast than than an AGCM driven by good SST?

No, (AGCM+1D ocean), the bias of the mean SST is too large by

AGCM+1D ocean, the resolved shorter-time scale flux is not

good enough.

Yes, (3-D OAGCM) only the bias from the SST mean state by the

coupled model can be overcomed by a better resolving the air-sea

interaction with time scale short than the mean state.

1. For example, if the rainfall by diurnal cycle and Mudian-Jullian

Oscillation (MJO) can be better resolved by the coupled model.

2. 𝐿𝐸 = 𝑤 ∙ 𝑞 = 𝑤 ∙ 𝑞 + 𝑤′ ∙ 𝑞′𝑚𝑜

+ 𝑤′ ∙ 𝑞′𝑚𝑗𝑜

+ 𝑤′ ∙ 𝑞′ℎ

Extended forecast: > 7d

B.J. Tsuang, Mong-Ming Lu

45 days of 500hPa GPH Anomaly Corr. Coef.

Coupled OAGCM

ECHAM/SIT/TIMCOM

(EHTW ESM), T106

More skill in the tropics: 200 hPa U comp. ACC for Tropical

(20oN~20oS)

(AMIP daily SST)

(AMIP daily SST)

(AGCM+3D Ocean)

(AGCM+1D Ocean)

Conclusion

Diurnal SST and MJO can be better captured by an OAGCM

with better treatment of fine vertical resolution in the upper

10m, co-allocated grid of AGCM and OGCM, exchange data

at ~40 min interval, and good vertical diffusivity scheme.

Precipitation can be better simulated by such an OAGCM

framework than AGCM driven by perfect SST