roperties of vegetation 
>and XS3 is best suited 
FOR ATMO- 
>N OF DUST 
raber (1993)) assumes 
e. These particles are 
two components: the 
ochastic nature of the 
cle hits the ground, a 
ned to be identical to 
’ the same process of 
e random walk model 
■o accelerations slower 
xactly. Secondly, the 
ter depletion of a dust 
Highway close to the 
ition being 40 x 40 m J . 
rrain is relatively fiat, 
d dispersion modeling 
rements was obtained 
e model domain. The 
tic winds of a nearby 
avitational settling, a 
oriented towards the 
t 5 different distances 
counted by means of 
esults show that the 
>out 7 pm. 
etermined during the 
tion will be shown in 
CTRAL DA- 
SPOT 
') sensors, HRV1 and 
sensitive in the three 
ed, 0.79 - 0.89pm). 
51 - 0.73 pm). The 
iunta to 20 m for the 
sufficiently well with 
: acquired only upon 
liable. Indeed, there 
d for our purposes is 
the local time zone), 
i Alaska Pipeline are 
ed as rivers or lakes, 
r lit 
■ 
- 
, (: i 
j 5 S 
ALASKA 
¡S-J 
Figure 1: Geographic location of the Toolik Lake area. SPOT scene, sensed in the HRV bands XS1 (0.50 - 
0.59 pm) and XS3 (0.79 - 0.89 pm) 
3.2 Method for deriving the dust load from the sensor signal 
This method allows to convert the raw multispectral image data of the HRV instrument to an array of dust load 
values which can be compared with the results of the stochastic model given in Sec. 2. The output signal of 
the satellite sensor is strongly correlated with the reflectance of the sensed object at ground level. Reflectance 
patterns having a finer resolution than the sensor’s instantaneous field of view (IFOV) are averaged over 
the IFOV. Including the reflectances of pure dust and pure vegetation (e.g. from image sections comprising 
sufficiently large areas of the respective land cover) the proportion of the area covered with dust can be derived 
for each image pixel. This proportion is a function of dust load, the correlation depending strongly on the 
dust particle size distribution which is known from the experiment. Hence, it is possible to derive the dust 
load from the signal of the satellite’s sensor. The conversion is done by the following steps: 
a) calibration of the digital counts X(x, y) => apparent reflectance p'(x,y) at satellite level 
b) geometric correction ^ apparent reflectance f/(x,y), adjusted to the model grid 
c) atmospheric correction =»true reflectance p(x,y) at ground level 
d) calculation of the area proportion f{x,y) on the basis of p(x,y) 
e) calculation of the dust load u(x, y) on the basis of f(x, y) and the size distribution of the dust particles 
a) Calibration 
The apparent spectral radiance V [PVm -J /im _1 sr“ 1 ] is given as a linear function of the digital count X 
according to L' = X/At , ¿=1 to 3 being the index of the HRV bands XS1, XS2 or XS3. The calibration 
coefficient Ak can be calculated on the basis of the absolute gain and the instrumental gain factor transmitted 
ffi a leader file together with the image data (for details see SPOT (1987) where the At’» are updated every 6 
months). In this document an error of 10% for At is stated. Table 1 lists the values for the instrument HRV1 
which scanned the investigated scene. In the following, the average of the two values reported for March 20 
and September 20, 1987 was used for each band. 
The apparent reflectance p' is defined as 
cos z Ei 
where z is the solar zenith angle and Eo = Eo(\ c )\Wm~ 
at the effective center wavelength X c of the HRV bands XS1, XS2 and XS3. 
I is the extraterrestrial solar spectral irradiance