The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
Image 
Data 
GMT 
Satellite 
Latitude ( №) 
Longitude( E°) 
1 
2006/04/16 
3:50 
Terra 
35-50 
90-125 
2 
2006/04/16 
5:30 
Aqua 
35-50 
95-125 
3 
2006/04/16 
14:55 
Terra 
30-45 
90-125 
4 
2006/04/17 
2:55 
Terra 
30-50 
90-125 
Table 1 Four images used in this study 
3. METHOD AND ANALYSIS 
When the dust storm happens, lots of dusts particles get together 
to form a dust layer. The thick dust layer can absorb and reflect 
surface radiation and solar radiation, emitting and radiating at 
the same time, so the values of sensors change. Among the 36 
channels of MODIS, the visible and near infrared channels are 
used to measure objects’ reflection while the thermal infrared 
channels are used to measure objects’ brightness temperature. 
Comparing spectrum characteristic among dust, ground and 
cloud, some significance characteristics can be found: Cloud 
has high reflection but low brightness temperature; ground has 
low reflection but high brightness temperature; the reflection 
and brightness temperature of dust storm are between the 
two.So in this paper, based on the reflection and brightness 
temperature of dust storm, the discriminate functions have been 
constructed to extract the scope of dust storm and estimate the 
dust storm intensity grades. 
2.1 The Scope of Dust Storm Extraction 
The scope of dust storm extraction is that dust storm is 
separated from cloud, snow, ground, etc and the boundary of 
dust storm is depicted accurately (Fan Yida,Shi Peijun,Wang 
Xiushan,2002). Since the 31, 32 bands of MODIS are in the 
thermal-IR window, absorption by other atmospheric gases is 
negligibly small and dust has a higher emissive at band 32 than 
at band 31 ,the BTD (brightness temperature difference) between 
the band 31 and band 32 can be used to detect the dust storm. 
Four interested regions: sand, cloud-sand, cloud and land, are 
chosen (see figure lj.From figure 2, the threshold -IK of BTD 
can be used to identify sand from other objects. But if the 
threshold of -IK is used, some dust storm areas cannot be 
detected. In order to extract the scope of dust storm accurately, 
the threshold of OK is used. In this way, ground and cloud 
should be separated farther. However, the infrared radiance is 
primarily sensitive to the upper cirrus cloud layer, especially 
when the upper-layer cirrus are thick, the BTD cannot be 
possible to detect dust under cirrus cloud, thus other bands 
should be considered 
Figure l.The MODIS true color image on April, 17, 2006, 
divided into four regions: sand, cloud-sand, cloud and land. 
Figure 2. Comparison of the brightness temperature difference 
(BTD) between the 31 band and the 32 band for four regions: 
sand, cloud-sand, cloud and land. 
At daytime, based on the reflection of different objects, ground 
and cloud can be separated. The reflection of dust storm was 
similar to that of ground, increasing with the wavelength. 
According to the spectral characteristic, the reflection of cloud 
and snow reaches the maximum at the third band of 
MODIS(0.459pm-0.479p.m),but the minimum at the seventh 
band(2.105pm -2.155pm).Based on the difference among the 
reflection of dust storm cloud > snow, the normalized 
difference dust index (NDDI) (John J. Qu, Menas Kafatos,2006) 
can be used to eliminate the influence of cloud and snow 
effectively. The NDDI can be written as 
NDDI = 
(¿7-63) 
(¿7 + 63) 
where b3,b7 are reflection at the third band and the seventh 
band of MODIS.