Transcript Cloud Radiation Forcing in Asian Monsoon Regions Simulated by

Cloud Radiative Forcing in
Asian Monsoon Region
Simulated by IPCC AR4
AMIP models
Jiandong Li, Yimin Liu, Guoxiong Wu
State Key Laboratory of Atmospheric Science and
Geophysical Fluid Dynamics (LASG), Institute of
Atmospheric Physics (IAP), Chinese Academy of
Sciences, Beijing
UAW2008, Tokyo, Jul. 2, 2008
Outline
 Study motivation
 Data and methodology
 Analysis results
– Climatology of CRF* in AMR*
– Annual cycle of CRF around East Asia
 Conclusion
CRF*: Cloud Radiative Forcing
AMR*: Asian Monsoon Region
Study motivation (1)
IPCC AR4, 2007
 Clouds are important modulator of climate. The
concept of CRF has been used extensively to study
the impact of clouds on climate (Ramanathan, 1989).
 In the current climate, clouds exert a cooling effect
on climate corresponding to the global warming.
Meanwhile, Cloud feedbacks remain the largest
source of uncertainty in climate sensitivity estimates
(IPCC AR4, 2007).
Study motivation (2)
Bin Wang, 2002
 There exists significant difference for circulation,
1) Could most AGCMs from IPCC AR4 reproduce the
precipitation and cloud radiative process in different
basic
of CRF in AMR?
areas
of features
AMR.
2) What are the main deficiencies for CRF simulation?
 So far most AOGCMs do not simulate the spatial or
intra-seasonal variation of monsoon precipitation
accurately.
Data
 ERBE data (Barkstrom et al, 1990)
– Monthly data from 1985 to 1989
– Resolution is 2.5°×2.5°and uncertainty is ±5 Wm-2
 IPCC AR4 AMIP data
– Monthly data from 1979 to 1993
– Interpolation into ERBE grids
 CMAP precipitation
Methodology
 CRF
– Long-wave CRF
– Short-wave CRF
 Study area
LWCF  Fircl  Fir
SWCF  S  cl  


– 60-150°E and 0-50°N including main AMR
 Area mean bias and RMSE
 Taylor diagram analysis (Taylor, 2001)
IPCC AR4 AMIP models
CCSM3
NCAR, USA
T85L26
CNRM-CM3
Centre National de Recherches
Meteorologiques, France
T42L45
GFDL-CM2.1
NOAA/Geophysical Fluid
Dynamics Laboratory, USA
2.0°×2.5°L24
GISS-ER
NASA/Goddard Institute for
Space Studies, USA
4°×5°L20
INM-CM3.0
Institute for Numerical
Mathematics, Russia
4°×5°L21
IPSL-CM4
Institute Pierre Simon Laplace,
France
2.5°×3.75°L19
MRI-CGCM2.3.2
Meteorological Research
Institute, Japan
T42L30
MPI-ECHAM5
Max Planck Institute for
Meteorology, Germany
T63L31
MIROC3.2(medres)
CCSR/NIES/FRCGC, Japan
T42L20
UKMO-HadGEM1
Hadley Center, Britain
1.25°×1.875°L38
Analysis results (1)
Climatology of CRF* simulated
by AMIP models in AMR*
0-50°N , 60-150°E
Observational climatology of CRF in AMR
DJF
1)
JJA
2)
In observation data, there is a near cancellation
between LWCF and SWCF at TOA in tropical deep
convective regions. However, the net CRF is very
large in AMR (M.Rajeevan et al, 2000), and the
SWCF in the East of TP is very strong(Yu et al,
2001, 2004).
What about the performance of model?
LWCF by AMIP models in DJF
GFDL-CM2.1

Four
models reproduce
MIROC3.2(medres)
the weak LWCF
MRI-CGCM2.3.2
between
Indian Byland
and Bengal Bay.
UKMO-HadGEM1

No model capture the
strong LWCF over TP*.

Positive LWCF
simulated by most of
models is lower than
observation between
East China and Japan.
TP*: Tibet Plateau
SWCF by AMIP models in DJF
GISS-ER

Four
models capture
MPI-ECHAM5
the strong SWCF in
MRI-CGCM2.3.2
East
of TP in DJF .

UKMO-HadGEM1
MME10
failed to
reproduce the SWCF in
East of TP, which is
caused by the biases of
most of models in this
region.
LWCF by AMIP models in JJA

In active convective
regions, the location
and intensity of LWCF
by most models have
larger biases.
SWCF by AMIP models in JJA

In active convective
regions, the location
and intensity of SWCF
by most models have
larger biases.

The same difficulty of
CRF simulation also lies
in Southwest of China
downstream of TP.

The LWCF and SWCF
simulated by AMIP
models are correlated
well with simulated
rainfall.
Rainfall by AMIP models in JJA

Compared to the spatial
pattern of simulated
CRF, particularly SWCF,
simulated rainfall shows
the similar spatial
pattern. This is more
clear in MME10 results
The relationship between CRF and
rainfall in AMR in JJA
EASM
ISM
WNPSM
ISM:
5-20°N，70-100°E
EASM:
5-20°N，110-140°E
WNPSM: 20-35°N，100-130°E
Correlation between rainfall and LWCF
R=0.585
R=0.455
R=0.143
R=0.242
R=0.625
R=0.889
Correlation between rainfall and SWCF
R=-0.673
R=-0.408
R=0.342
R=-0.742
R=-0.826
R=-0.493
The spatial patterns of observational and simulated CRF
have good correlation with corresponding rainfall, which
very likely indicates two questions as following:
1. Generally, the simulated rainfall is directly connected
with cumulus parameterization process in model,
which affects rainfall, cloud physical process and
CRF(Zhang, 2006). Hence, the larger biases of CRF is
very likely to related to cumulus parameterization
scheme in model.
2. Many studies (Bin Wang et al, 2004, 2005) showed that
AGCMs are unable to realistically reproduce AsianPacific summer monsoon rainfall due to neglecting the
atmospheric feedback on SST , but AOGCMs have
better performance in rainfall simulation, SST and their
variability in AMR.
a. Comparison CRF by coupled model with that by AGCM
b. Relation between CRF and rainfall in coupled models
Area mean bias and RMSE
LWCF
SWCF
DJF
JJA
Units: Wm-2
Taylor diagram analysis for CRF
1)
2)
3)
There are large diversity and biases of CRF by models.
The diversity and biases of SWCF is larger than that of LWCF
especially in JJA.
GFDL-CM2.1, MPI-ECHAM5, UKMO_HadGEM1 and MME10
perform well in CRF simulation.
Analysis results (2)
Annual cycle of CRF* simulated
by AMIP models around EA*
EA*: East Asian
0-50°N , 100-145°E
Annual cycle of observational CRF
 In tropical area (south to 20°N) ，the variation of CRF
is consistent with that of rain season。
 In East of TP (between 25°and 40°N), stronger CRF
appears since February and lasts until late May, when
CRF evolves north with the rain season.
Annual cycle of CRF by AMIP models
GISS-ER
GFDL-CM2.1
MPI-ECHAM5
UKMO-HadGEM1
Conclusion
 There still exists a lot of difficulty in simulating the CRF in
AMR. Our study shows that the lee slide of TP in DJF
and JJA and active convective regions in JJA, such as
Bengal Bay, are the major bias regions.
 Further analysis indicates the biases and diversity of
SWCF are larger than that of LWCF. As a whole, GFDLCM2.1, MPI-ECHAM5, UKMO-HadGEM1 and
MIROC3.2(medres) perform well in CRF simulation in
AMR.
 It is suggested that strengthening the study of physical
parameterization involved in TP, improving cumulus
convective process and ameliorating model experiment
design will be crucial to the CRF simulation in AMR.
Thank you
谢谢
[email protected]
Centered RMSE
Units: Wm-2
Annual cycle of rainfall by AMIP models