The impact of cloud water schemes on seasonal prediction

1
The impact of cloud water schemes on seasonal
prediction
Akihiko Shimpo and Masao Kanamitsu
Experimental Climate Prediction Center (ECPC),
Scripps Institution of Oceanography (SIO), UCSD,
La Jolla, CA
ashimpo_at_ucsd.edu
Introduction Clouds are one of the most
uncertain components in climate models. The
model results are known to be very sensitive to
the parameterization of cloudiness. In this
study, relationships between cloud amount, cloud
water and relative humidity are compared between
observation and model simulations. The
observation is taken from monthly mean cloud
amount and cloud water path from ISCCP D2 data
and relative humidity from NCEP/DOE reanalysis 2
data. The difference of the zonal mean
temperature cross section in the long term
simulation with different cloud schemes are also
compared and possible relation with cloudiness
parameterization is studied.
From observation / reanalysis From scatter
diagrams for relative humidity and cloud amount
(Fig.1a), positive correlations are seen,
especially middle cloud. But the range of cloud
amount is wide for a given relative humidity.
Negative correlation is seen for low cloud over
land (Fig.1d). This might be caused by the
satellite-based observations, in which low clouds
can be seen when they are not overlain on higher
clouds (ex. Hahn et al., 2001). The
relationships between relative humidity and cloud
water (Fig.1b) for each cloud are similar to
those between relative humidity and cloud amount.
It shows that cloud water paths spread for a
given relative humidity. The correlation between
cloud water and cloud amount (Fig.1d) is higher
than that between relative humidity and cloud
amount, especially for high cloud, the range of
cloud amount for a given cloud water path is
narrow, while that of cloud amount for a given
relative humidity is wide.
From simulations From CNTL simulation
(Fig.2a), in which cloud amount is derived from
only relative humidity, the scatter diagrams show
that relative humidity and cloud amount correlate
for high and middle clouds. The range of cloud
amount for a given relative humidity is narrower
than that of observation and reanalysis. Negative
correlation between relative humidity and low
level cloud amount is found. These clouds are
calculated as clouds under temperature inversion
layer over ocean, which include marine stratus
and shallow cumulus clouds. From ZCR simulation
(Fig.2b), in which cloud water is predicted and
cloud amount is derived from relative humidity
and cloud water content, the relationship between
relative humidity and cloud amount is similar to
that of CNTL simulation. Negative correlation for
low cloud is also found like CNTL because
temperature inversion type clouds in CNTL were
added to the ZCR simulation to represent marine
stratus. Simulated cloud water path for middle
cloud is similar to observation, but for high
cloud is smaller than that of observation. One of
reasons might be the setting of critical value of
snow auto conversion. Cloud water path for low
cloud is more concentrated for high relative
humidity than that for observation. From IS
simulation (Fig.2c), in which cloud water and
cloud amount are predicted, cloud amount and
cloud water path to relative humidity are more
sensitive than that from observation and
reanalysis. The range of cloud amount to cloud
water path is wider than that of
observation. Sensitivity map was plotted with
regression coefficients of d(C)/d(RH) and
d(C)/d(CWP) from observation/reanalysis and 3
simulations (Fig.4). It shows that 3 simulations
have stronger sensitivity of cloud amount to
relative humidity than observation/reanalysis. Si
mulated zonal mean temperature distributions are
similar between CNTL and ZCR, while IS simulation
showed different distribution (Fig.3).
Data and forecast model The International
Satellite Cloud Climatology Project (ISCCP) D2
series data (Rossow and Schiffer, 1999) was used
for cloud amount and cloud water path. Cloud
amounts were provided as 3 cloud types, high,
middle and low clouds. Cloud water paths were
calculated for 3 cloud types, which are deduced
from 15 cloud types in the original datasets.
NCEP/DOE reanalysis 2 data (Kanamitsu et al.,
2002) was used to estimate relative humidity. 3
layer relative humidity was derived from maximum
relative humidity in each 3 layer. All data were
global monthly mean, 2.5 degree lat/lon
resolution. Calculations were done in the region
between 60S and 60N because cloud detection
errors are expected to be larger in the polar
regions (Rossow and Schiffer, 1999). Data from
Mar/2000 to Feb/2001 were used. The results are
very similar when data for longer period are
used. ECPC GSM was used as forecast model.
Resolution was T62L28. 3 types of AMIP-type runs
were run from the 1st of Jan 2000 for 15months
with different cloud water and cloud schemes (
table.1 ). Calculations were performed on the
simulation from Mar 2000 to Feb 2001.
Random-maximum overlapping assumption was applied
to calculate 3 types of clouds in the models
(Chou et al., 1999).
(d)RH-C (land)
(a)RH-C
(b)RH-CWP
(c)CWP-C
high
high
high
high
C
C
C
CWP g/m2
middle
middle
middle
middle
C
C
C
CWP g/m2
low
low
low
low
CWP g/m2
C
C
C
RH
RH
RH
CWPg/m2
Conclusion From ISCCP cloud amount and cloud
water observations and Reanalysis-2 relative
humidity, it is found that the correlation
between cloud water and cloud amount is higher
than that between relative humidity and cloud
amount. This suggest that there is a distinct
advantage in predicting cloud water than simply
diagnosing from relative humidity. It also seems
to suggest that the combination of relative
humidity and cloud water may not be necessary to
obtain cloud water. Negative correlation was
found between relative humidity and low cloud
amount over land, but this is most likely due to
the observation from satellite. Negative
correlation between relative humidity and low
level cloud amount was found in ZCR and CNTL
experiments. These clouds are diagnosed from
boundary layer strength and height and may not be
realistic. The range of cloud amount for a given
humidity in the forecast models is narrower than
that in observation, while the range of cloud
amount for a given cloud water is wider in the
model than observation. The sensitivity of cloud
amount to relative humidity and to cloud water is
different for different cloud schemes. The
sensitivity of ZCR is similar to CNTL. IS
cloudiness is more sensitive to relative humidity
than observation and ZCR and CNTL. Simulated
zonal mean temperature distributions are similar
between CNTL and ZCR, which might be explained by
the fact that the relation between cloud amount
and relative humidity between the two models are
similar. On the other hand, IS simulation, which
has different sensitivity to relative humidity
and cloud water than other two schemes, showed
different temperature distribution.
Fig.1 Scatter diagrams of (a) relative humidity
() vs. cloud amount (), (b) relative humidity
() vs. cloud water path (g/m2), (c) cloud water
path (g/m2) and cloud amount () and (d) same as
(a) except for over land only. for (top) high,
(middle) middle and (bottom) low clouds from
observation and reanalysis. For the area
(a)(b)(c) 0-360,60S-60N and (d)60E-120E,
30N-60N, and the period Mar/2000 Feb/2001.
Cloud water Cloud amount Equation of cloud amount
CNTL (non) Diagnostic RH (Slingo and Slingo, 1991)
ZCR Prognostic (Zhao and Carr, 1997) Diagnostic RH / CW (Randall, 1995)
IS Prognostic (Iacobellis and Somerville, based on Tiedtke, 1993) Prognostic (Iacobellis and Somerville, based on Tiedtke, 1993)
(a)RH-C
(b)RH-CWP
(c)CWP-C
CNTL
CNTL
high
middle
C
low
ZCR
ZCR
ZCR
ZCR
C
C
CWP g/m2
IS
IS
IS
IS
CWP g/m2
C
C
Table.1 cloud water and cloud schemes
RH
RH
CWPg/m2
Fig.2 Scatter diagrams of (a) relative humidity
() vs. cloud amount (), (b) relative humidity
() vs. cloud water path (g/m2) and (c) cloud
water path (g/m2) and cloud amount () from (top)
CNTL, (middle) ZCR and (bottom) IS simulation.
For the area 0-360,60S-60N and the period
Mar/2000 Feb/2001. Red, green and blue dots are
for high, middle and low clouds.
Fig.3 Zonal mean simulated temperature
difference from R-2 (K) for 1 year average from
Mar/2000 to Feb/2001. (top) CNTL, (middle) ZCR
and (bottom) IS simulation.
()/(g/m2)
Z
high
middle
low
1.0
C
C
C
References Chou, M.-D., K.-T. Lee S.-C. Tsay,
and Q. Fu, 1999 Parameterization for cloud
longwave scattering for use in atmospheric
models. J. Climate, 12, 159-169. Hahn, C. J., W.
B. Rossow, and S. G. Warren, 2001 ISCCP cloud
properties associated with standard cloud types
identified in individual surface observations. J.
Climate, 14, 11-28. Kanamitsu, M., W. Ebisuzaki,
J. Woollen, S.-K. Yang, J. J. Hnilo, M. Fiorino,
and G. L. Potter, 2002 NCEP-DOE AMIP-II
Reanalysis (R-2). Bull. Amer. Meteor. Soc., 83,
1631-1643. Randall, D. A., 1995 Parameterizing
fractional cloudiness produced by cumulus
detrainment. Workshop on cloud microphysics
parameterizations in global atmospheric
circulation models, Kananaskis, Alberta, Canada,
23-25 May 1995, WCRP-90, WMO/TD-No.713, 1-16.
Rossow, W. B., and R. A. Schiffer, 1999
Advances in Understanding clouds from ISCCP.
Bull. Amer. Meteor. Soc., 80, 2261-2287. Slingo,
A., and J. M. Slingo, 1991 Response of the
National Center for Atmospheric Research
Community Climate Model to improvements in the
representation of clouds. J. Geophys. Res., 96,
15,341-15,357 Tiedtke, M., 1993 Representation
of clouds in large-scale models. Mon. Wea. Rev.,
121, 3040-3061. Zhao, Q., and F. H. Carr, 1997 A
prognostic cloud scheme for operational NWP
models. Mon. Wea. Rev., 125, 1931-1953.
0.5
I
Z
I
I
Z
0
0
0.5
1.0
()/()
Fig.4 Sensitivity map for cloud amount to
relative humidity and cloud amount to cloud water
path from observation/reanalysis and simulations.
Circles are for observation/reanalysis and
initial letters are used for simulations. Red,
green and blue mean high, middle and low clouds.
RH
RH
CWPg/m2