Retrieval of cloud parameters from the new sensor
generation satellite multispectral measurement
F. ROMANO and V. CUOMO
ITSC-XII Lorne, Victoria, Australia
Main Object:
Improvement using AMSU data and high spectral
resolution data (IASI and IMG)
in
•cloud detection
•cloud clearing
•cloud forcing radiative
•retrieval of cloud parameters
ITSC-XII Lorne, Victoria, Australia
TOP CLOUD HEIGH
The cloud radiative forcing the simple difference
between cloud and cleared radiances can be
useful in the estimation of top cloud height.
Accurate cloud clearing and cloud detection
packages are necessary to estimate cloud
radiative forcing.
ITSC-XII Lorne, Victoria, Australia
IASI Cloud Clearing Scheme
AMSU Data
IASI/AMSU
REG. COEF.
Synthetic IASI Data
IASI Data
Clear IASI FOV in the Box
Variogram
Kriging interpolation
Cleared IASI data and errors
ITSC-XII Lorne, Victoria, Australia
Cloud Mask
IASI Cloud Clearing Test
DATASET
•R. Rizzi’s Cloudy Dataset (CDS)
•M. Madricardi’s AMSU Brightness Temperature Data set
ITSC-XII Lorne, Victoria, Australia
CLOUD CLEARING RESULTS
Results, on the basis of IASI simulated data
and ATOVS data,
show that
the root mean square error of the Kriging
clear brightness temperatures estimates is
well below 1°K for any IASI or HIRS
channels.
ITSC-XII Lorne, Victoria, Australia
IASI/AMSU CLOUD DETECTION
AMSU data have been used in order to have some
information about the clear radiance field.
The test based on AMSU data allow to detect all the
overcast FOVs and high cloud.
Low clouds, thin cirrus clouds and partially FOVs
cloudy is detected using some tests based on IASI or
IMG signatures.
ITSC-XII Lorne, Victoria, Australia
IASI/AMSU CLOUD DETECTION
VALIDATION
•R. Rizzi’s Cloudy Dataset (CDS)
•M. Madricardi’s AMSU Dataset
Data used
FOVs
number
91200
Clear FOVS
detected cloud
Cloud FOVS
detected Clear
FOVS detected
Exactly
2418 ( 2.6 % )
2485 ( 2.7 % )
86297 ( 94.6 % )
ITSC-XII Lorne, Victoria, Australia
IMG/AMSU CLOUD DETECTION
VALIDATION
•MEASURED IMG DATA
•COLLOCATED AMSU/MSU DATA
Data used
Clear FOVs
FOVs
detected cloud
number
580
12 (2.06%)
ITSC-XII Lorne, Victoria, Australia
Cloud FOVs
detected Clear
26 (4.48%)
FOVs detected
Exactly
542 (93.45%)
CO2 slicing method
CO2 slicing has been extensively used to
retrieve cloud top pressure and cloud
effective emissivity. The radiance from a
partially cloudy air column region can be
written as:
R   R
cloud
 (1   ) R
clear
Using Radiative Transfer Equation and after
simple operation we obtain:
R  R
clear
 
pc
   ( p)dB
ps
ITSC-XII Lorne, Victoria, Australia
CLOUD TOP HEIGHT : Single Cloud Level
For two close spectral channels and viewing the
same FOV can be written as:
pc
 R
clear
R
R  R
1
1
clear
2
2
     ( p)dB
1


1
2
( p )dB2
ps
pc
2
 
ps
ITSC-XII Lorne, Victoria, Australia
1
In order to apply the CO2 slicing technique to
IASI data, it is necessary to select the best
pairs of frequency to be use in the cloud top
retrieval.
The method used, select all the channel in the
(700 – 753 cm-1) absorption band whose
weighting functions peak between 200 mb and
900 mb.
It use all the possible combinations of those
channels, with the first channel in the pair
always associated with the lower wavenumber.
ITSC-XII Lorne, Victoria, Australia
Then apply the CO2 slicing technique to retrieve the
cloud top heights, using all the selected channel pairs.
Finally selected the number of pairs that best satisfy
the radiative transfer equation for all the spectral
channels.
For each FOV the pairs number of different solutions
found are used to evaluate a cost function:
N
    2 i
 i 
ITSC-XII Lorne, Victoria, Australia
R
1
 R
clear
i
i
   
pc
i
ps
i
( p)dBi
The solution associated to the smallest values of
cost function are averaged to determine the cloud
top height.
Increasing the number of channel pairs used in
average causes a improvement in the accuracy of
the top cloud height retrieval.
At the end the algorithm select 36 pairs of
channels.
ITSC-XII Lorne, Victoria, Australia
When the cloud height is known, the cloud effective
emissivity has been estimate using the infrared
window channel data (893 cm-1) by means from the
follow relation:
 
ITSC-XII Lorne, Victoria, Australia
R R
w
B
clear
w
(T ( pc ))  Rw
clear
w
To examine the performance of the top cloud
height and effective emissivity algorithm, the
numerical simulations of clear and cloudy
radiances were carried out using synthetic data
from LBLRTM.
Additional instrumental noise were simulated
according to IASI specification noise level.
ITSC-XII Lorne, Victoria, Australia
Radiances are simulated for ten cloud-top
pressure (900 mb, 850 mb, 830 mb, 800 mb,
730 mb, 750 mb, 730 mb, 700 mb, 650 mb, 600
mb)
for four cloud amounts (0.1, 0.5, 0.7, 1)
and for the U.S. Standard climatological profile,
for all the frequency used in the algorithm
measured
ITSC-XII Lorne, Victoria, Australia
CLOUD TOP HEIGHT : Two Cloud Level
The cloud forcing when you consider two level
cloud (upper and lower cloud) can be expressed
as:
R  R
clear
pcl
pch
ps
ps
  l  l (1   h  h )    ( p)dB   h h    ( p)dB
From all the solutions we select these best satisfies the radiative
transfer equation for all spectral channels.
ITSC-XII Lorne, Victoria, Australia
In the same way we defined a cost function and the
solutions associated to the smallest values are averaged to
determine the upper and lower cloud top height.
Increasing the channel pairs number used in average
causes a improvement in the accuracy of the top cloud
height retrieval.
At the end the algorithm select 44 pairs of channels
ITSC-XII Lorne, Victoria, Australia
Cloud Top height Package Validation
The top cloud height package has been apply at the simulated
CDS-IASI cloud data.
Between the 91200 FOVs 1010 FOVs with a single cloud layer
and 605 FOVs with two cloud levels have been selected.
The temperature profile and the profiles of atmospheric
transmittance for the spectral frequency have been calculated
as a function of top cloud top pressure.
The cloud top altitude is assumed to vary according to the 43
pressure discrete layering.
ITSC-XII Lorne, Victoria, Australia
Single Cloud Level
Cloud top pressure (mb )
1000
a=4.70
b=0.99
r=0.99
sd=1.61
800
600
400
200
0
0
200
400
600
800
Retrieved Cloud Top Pressure (mb )
ITSC-XII Lorne, Victoria, Australia
1000
Two Cloud Levels
Cloud top pressure (mb )
400
a=-0.78
b=0.99
r=0.99
sd=1.56
300
200
100
0
0
100
200
300
Retrieved cloud top pressure (mb )
ITSC-XII Lorne, Victoria, Australia
400
Two Cloud Levels
1000
a=6.03
b=0.99
r=0.99
sd=1.77
Cloud top pressure (mb )
950
900
850
800
750
700
650
650
700
750
800
850
900
Retrieved cloud top pressure (mb )
ITSC-XII Lorne, Victoria, Australia
950
Conclusions and Future Work
AMSU data and high spectral resolution IASI and IMG
sounders data can greatly be improved retrieval cloud
parameters.
The cloud top pressure and the effective emissivity retrieval
are in good agreement with the true theoretical values.
The cloud top pressure retrieved packages will be extend to
multiple cloud levels.
Validation, based on radiosonde and meteorological radar,
will be extend to IMG and AIRS measured data.
ITSC-XII Lorne, Victoria, Australia