polarization channels on the principal plane of POLDER 
data taken on April 12, 1997. The principal plane passes 
over the “Kosa”-like cloud in the range of the scattering 
angle between 110 and 135 degrees. We can see from Fig.3 
that the degree of polarization at the wavelength 443nm 
greatly decreases in the range of the scattering angle less 
than 135 degree. 
We do not measure the concentration and optical properties 
of yellow sand dust at the ground stations. Therefore, it is 
not clear that the *Kosa"-like cloud as shown in Figs.1 and 
2 actually consists of yellow sand dust particles whirled up 
in the desert of the northern part of China.. 
3. LONG-RANGE TRANSPORT OF “KOSA” 
One method for finding the “Kosa” cloud on satellite-level 
data is to simulate the long range transport of yellow sand 
dust particles. We used the long range transport software: 
Mirage 3.0 developed by EDF and Aria technologies, 
France to obtain the distribution of yellow sand dust over 
East Asia. The Mirage 3.0 allows us to compute the 
diffusion and depletion of yellow sand dust as well as the 
trajectories of the air mass on the basis of meteorological 
data given by ECMWF (European Center for Medium 
range Weather Forecast). Unfortunately, we have only 
ECMWF data from March to April, 1995 and so we can 
not show the simulation results of the trajectories and 
concentration of yellow sand dust at the time when 
POLDER data were acquired. However, we consider that it 
will be possible to estimate the distribution of yellow sand 
dust on POLDER images over East Asia by using the 
ECMWF data in 1995. 
On April 8, 1995, the Kosa phenomenon was recognized 
at the meteorological station in Ishikawa, Japan, which is 
located at the side of the Sea of Japan. The concentration of 
the suspended particulate matter measured at ground 
stations in Ishikawa prefecture in Japan was high from 
April 8 to 9 (Report 1995) and its average value was 53 4 
g/m’. We also found that the high concentration of the 
suspended particulate matter was measured at ground 
stations on April 5 and 6. On the basis of this observational 
evidence, we determined the starting date and source 
regions for the transportation of yellow sand dust by 
referring to the weather chart. Since the low pressure 
appeared in the northern part of China on March 27- 28, 
1995, March 27 was chosen as the initial date for the 
numerical simulation, and Gobi (40° N, 103E) and 
Badain Jaran (42N, 100E) deserts were selected as source 
regions. The Kosa particles were released at several 
altitudes every hour from 00 GMT March 27 to 00 GMT 
April 9. The size of yellow sand particles was assumed to 
be 1 4 m. Fig.4 shows the distribution of concentrations 
(larger than 53 44 g/m’) of yellow sand dust over East Asia 
on April 8 in the case that “Kosa” particles were released at 
the altitude of 750hpa in Gobi and Badain Jaran deserts. 
Fig.5 shows the distribution of yellow sand dust over East 
Asia on April 6 in the case that “Kosa” particles were 
released at the altitude of 650hpa in Gobi and Badain Jaran 
deserts. As seen from Fig.5, there is a distinct belt of high 
concentrations of yellow sand dust, which extends from the 
main land of China to Korea. We imply from the results of 
this simulation that the thick Kosa-like clouds that are 
visible in Figs.1 and 2 will consist of yellow sand particles. 
We can also see from Fig.4 that almost all part of East Asia 
is covered with the hazy Kosa layer (gray regions in Figs.4 
and 5). 
4. POLARIZATION PROPERTIES OF KOSA 
As seen from Fig.3, the polarization of 443nm channel is 
smaller than that of other polarization channels in the Kosa 
cloud. To interpret this, we computed P,(=-P,2/P;1), the 
linear polarization for single scattering of incident 
unpolarized light in terms of Mie scattering formula, where 
Py; and Py; are components of phase matrix (Hansen 1974). 
The components of phase matrix are sensitive to the shape 
of the size distribution of Kosa particles and so the size 
distribution of the power law having the power -4 was 
assumed in this study. And the refractive index of Kosa 
particles was assumed to be 1.55, because in Japan the 
aerosol refractive index varies 1.47 to 1.57 and its value 
increases in spring (Tanaka, 1983). Even if the size 
distribution model and refractive index are fixed, the value 
of P, depends on the smallest (r1) and largest (12) particle 
sizes in the numerical integration of the size distribution. 
The condition that the linear polarization at the 443 channel 
(Ps)a43 becomes smaller than that at the 865 channel (Ps)ges 
in the range of scattering angles between 110 and 140 
degrees was tested for various cases of rl and r2. As a 
result, it was shown that we have (P.)443 € (P,)sg; in the case 
of r1™~1um and r2~ Sum. There is a good agreement 
between the simulated particle sizes and particle sizes in the 
mass size distribution derived from Kosa events in 1981 
(Arao 1986). The mass size distributions of yellow sand 
dust obtained from Kosa events have a peak at the radius 2 
um. 
S. CONCLUSIONS 
We used the polarization measured from the POLDER 
sensor and the long range transport simulation of yellow 
sand dust to identify *Kosa clouds" on ADEOS/PODER 
images. As a result, we found that the linear polarization at 
the 443 channel becomes smaller than that at the 865 
channel in the range of scattering angles between 110 and 
135 degrees in the thick Kosa cloud, and that the size of 
Kosa particles is in the range from 1 jum to Sum. It was also 
shown from the long rangé transport simulation that the 
440 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
EAS €) AN M9