ally- 
ıtion 
with 
over 
dual 
sd to 
fast 
g of 
rt as 
zenic 
from 
001) 
rties, 
and 
cale. 
135, 
loud 
35 to 
are 
over 
is in 
] for 
hr 
or 
73K. 
us r. 
two 
rmal 
solar 
of 1, 
is a 
ithm 
)5)’s 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
incorporated new treatments on thermal emission and water 
vapor absorption in the atmosphere. A pre-computed numerical 
table (lookup table) was prepared to compare with satellite 
observed radiances, and it determined the best-matched 
combination from the comparison to be the retrieved cloud 
properties. To construct lookup table, they used a general 
radiative transfer (RT) code which solves RT with a combined 
discrete-ordinate-matrix-operator method (Nakajima and 
Tanaka 1986, 1988) and LOWTRAN-7 gas absorption model 
(Kneizys et al. 1988). About ancillary data for actual data 
analysis, NCEP/NCAR reanalysis data were used for input 
meteorological data such as humidity, pressure and temperature 
profiles, and the minimum value for the month was determined 
as the monthly ground albedo. Retrieved results were well 
validated with in situ aircraft measurements (Kawamoto et al. 
2001). 
This algorithm was applied to AVHRR (Advanced Very High 
Resolution Radiometer) GAC (Global Area Coverage) radiance 
data for 1985 - 1988 (NOAA-9) and 1989 - 1994 (NOAA-11). 
4-month analysis (April, July, October and December) was 
performed, and the annual mean was taken as their arithmetic 
average. As for calibration of sensor signals, the calibration 
constants from Rao and Chen (1994) were used for channel 1 
visible channel, and those from on-board internal blackbody for 
channel 3 near-infrared and channel 4 infrared channels. 
Potentially analyzed pixels were selected whose satellite zenith 
angles less than 25 degrees in order to avoid the effect of cloud 
inhomogeneity. According to Iwabuchi and Hayasaka (2002), 
that effect would be reduced for the most part when viewed 
from angles less than about 25 degrees. For efficient processing, 
target area was divided into 0.5-degree spatial segment and one 
segmented-box stores 100 pixels. We constructed this 
segmented data daily for the analysis period, and analyzed one 
pixel which had the median value of visible reflectivities among 
those classified as cloudy in each segmented box. 
3. Results and discussion 
1) The annual-mean cloud characteristics over China 
First, we begin with the annual features of cloud properties 
over target area in 1990. Fig.1 shows the annual mean of x. 
Generally x over the land is larger than that over the ocean. 
Especially large t is observed in the southern part of China. 
This cloud would be occurred mainly due to active convection. 
These general features are consistent with ISCCP (International 
satellite Cloud Climatology Project) statistics (Rossow et al. 
1996). Fig.2 shows the annual mean of r,. Unlikez, r, over land 
is smaller than that over the ocean. Particularly r, over the 
eastern and central parts of China is smaller, although that over 
the north and west part are larger. Fig.3 shows the annual mean 
of the integrated cloud droplet number N,. 
N, is estimated from x and r, with the assumed size distribution 
of log-normal. We find large N, areas for small r, and large T. 
Values of N, reported here are compatible with the preceding 
study (Han et al., 1998). These features on r, t and N, can be 
explained by Twomey effect. An idea was proposed that 
additional aerosol particles can decrease the cloud particle size 
and increase the cloud optical depth, with increasing cloud 
droplet number (Twomey, 1977). In general, acrosols are less 
over ocean, and more over land. The above phenomena are 
consistent with this idea. Note that cloud properties near the 
east coast are affected by continental airflow, that is, r, is 
smaller, while t and N, are larger over land compared to those 
of remote oceanic area. 
As stated above, patterns of cloud properties over land are not 
geographically uniform. Kawamoto et al. (2003) investigated 
their relationships with SO2 emission, which is location- 
775 
dependent. SO2 emission is more in highly populated coastal 
areas and inland industrial cities, and SO2 is known as a 
precursor gases for sulfate aerosols. They found good 
agreement between the two. 
  
  
0 5 10 15 20 
Fig.l The annual-mean of low-cloud optical depth 
  
6 8 5° 1 13] 15 16 18 
Fig.2 The annual-mean of low-cloud effective particle size 
  
(1/em?) 
0.00e+00 5.00e+06 1.00e+07 
  
Fig.3 The annual-mean of low-cloud droplet number