Diagnosis of Regime-dependent Cloud Simulation Errors

meatcologneInternet και Εφαρμογές Web

3 Νοε 2013 (πριν από 3 χρόνια και 5 μήνες)

116 εμφανίσεις

Diagnosis of Regime
-
dependent Cloud Simulation Errors

Hui Su, Jonathan H. Jiang, C. Zhai, V. S. Perun, J. T. Shen

Jet Propulsion Laboratory, California Institute of Technology

ISCCP at 30, New York, NY, April 22
-
25, 2013

with contributions from

A. Del Genio, L.S. Nazarenko, L.J. Donner, L. Horowitz, C. Seman, C. Morcrette,
J. Petch, M. Ringer, J. Cole, M. Mesquita, T. Iversen, J.E. Kristjansson, A.
Gettelman, L. Rotstayn, S. Jeffrey, J.L. Dufresne, M. Watanabe, H. Kawai, T.
Koshiro, T. Wu, E.M. Volodin, T. L'Ecuyer, J. Teixeira, and G.L. Stephens


Cloud distributions are closely coupled with large
-
scale
dynamic and thermodynamic conditions.


Cloud feedbacks are strongly correlated with large
-
scale
circulation changes.

Motivation

G
l
o
b
a
l
-
m
e
a
n

N
e
t

C
R
E

C
h
a
n
g
e

(
W
/
m
2
)

Corr=0.74, Slope=0.64 W
m
-
2

K
-

Corr=0.74, Slope=0.64 W m
-
2

K
-
1


ΔTs (K)

ΔCRE (W/m
2
)

Corr=0.74


Change of Hadley Cir. Strength (10
10
kg/s)

ΔCRE (W/m
2
)

Corr=0.77

Su et al. (in preparation)


Examine the relationships between clouds and their large
-
scale environmental conditions

Objectives


Mid
-
tropospheric velocity (
ω
500
)



Lower
-
level Divergence (LDV)


Sea Surface Temperature (SST)


Lower Tropospheric Stability (LTS)


Water Vapor Path (WVP)


Relative Humidity (RH)

Thermodynamic
Parameters

Dynamic
Parameters


Diagnose the source of cloud simulation errors

Su et al., JGR, 2013

Diagnostic Framework

Cloudiness in a given large
-
scale regime (V)

Tropical
-
mean cloudiness

Probability density function of a large
-
scale regime (V)

Difference between modeled and observed cloudiness in Regime V

Difference between modeled and observed pdf of Regime V

Parameterization
error

Co
-
variation

error

Large
-
scale

error

Su et al., JGR, 2013

CMIP5 Clouds Sorted By
ω
500



Models capture the dominant modes of tropical clouds in distinct large
-
scale regimes.



Large discrepancy in the model simulated vertical structures of clouds.

Three Components of Errors

Large
-
scale

error

Parameteri
-
zation Error

Co
-
variation
Error

Total Error

Tropical
-
mean Cloud Errors


Cloud parameterization errors dominate the total cloud errors on the tropical mean.


Uncertainty in satellite retrievals complicates quantitative assessment of model biases.


Model simulator CF agrees better with the satellite data than the model original CF.

Su et al., JGR, 2013

Blue: CWC errors relative to the CloudSat total CWC

Green: CWC errors relative to the CloudSat non
-
precipitating CWC

Red: Original cloud fraction errors

Black: CALIPSO
-
simulator cloud fraction errors

Using the ISCCP Cloud Simulator Results

Cloud Fraction Sorted by
ω
500



Models capture the dominant modes of tropical clouds in distinct large
-
scale regimes.



Large discrepancy in the model simulated cloud amounts and detailed structures.

Using the ISCCP Cloud Simulator Results

Cloud Fraction Sorted by
SST



The differences in low and middle cloud amounts over cold SST regimes are large.



Except CCCMA, other models simulate well cirrus clouds over warm SST regimes.

Using the ISCCP Cloud Simulator Results

Sorted by SST (thin clouds)



Models capture well thin cirrus clouds in the upper troposphere, but not in the free troposphere.

Using the ISCCP Cloud Simulator Results

Sorted by SST (medium thick clouds)



Differences between models and data in both cold and warm SST regimes are evident.

Using the ISCCP Cloud Simulator Results

Sorted by SST (very thick clouds)



Large discrepancy in the cold SST (subsidence) regimes.

Using the ISCCP Cloud Simulator Results

Cloud Fraction Sorted by LTS



Models tend to represent well the relationship between LTS and low/middle clouds, but
the vertical extent of low/middle clouds varies from model to model.

Conclusions


The total cloud errors are primarily contributed by the cloud
parameterization errors.


The contributions of large
-
scale errors are secondary in the tropical
mean but can be substantial in specific large
-
scale regimes.


Using ISCCP simulator outputs, we find that model performance
varies across cloud optical depths and in different large
-
scale
regimes.

“Continue working on it,
until it breaks.”


Bill Rossow