1078 
of the spectral band of the sensor. However, it turned out that due to the spectral properties of vegetation 
and dust as well as due to the limited digital resolution of the signal, the near infrared band XS3 is best suited 
for deriving the dust load. 
2 OUTLINE OF THE RANDOM WALK MODEL FOR ATMO 
SPHERIC DIFFUSION AND DRY DEPOSITION OF DUST 
The random walk approach of simulating air pollution dispersion (Lamprecht and Graber (1993)) assumes 
that the polluting material is represented by fictitious particles released at the source. These particles are 
individually moved downwind. The relevant wind velocity vector is determined by two components: the 
average wind velocity and the random turbulent fluctuations of the air. Due to the stochastic nature of the 
air turbulence, each individual particle will follow its own trajectory. When a particle hits the ground, a 
certain probability for its dry deposition is taken into account. 
For gaseous diffusion the turbulence characteristics of fluid particles are assumed to be identical to 
those of the surrounding air. Heavy particles are transferred in the atmosphere by the same process of 
turbulent diffusion except that a few minor modifications are necessary to extend the random walk model 
for inertia-less particles: Firstly, due to the greater inertia, a heavy particle responds to accelerations slower 
than a fluid particle and hence does not follow direction changes of fluid elements exactly. Secondly, the 
earth gravity causes an additional downward velocity which contribute to an overall faster depletion of a dust 
plume. 
The above model was applied to a domain which is located along the Dalton Highway close to the 
Toolik Lake area (see Fig. 1). The size of this domain is 2.4 x 2.8 km 2 , with the grid resolution being 40 * 40 m 3 . 
The landscape is slightly slanted along the slope of a small hill. Apart from that the terrain is relatively flat, 
so that detailed topographical features were neglected for the deposition calculation. 
The knowledge of the mean windfield is of primary importance for transport and dispersion modeling 
of air pollution species. This information together with shielded air temperature measurements was obtained 
from a meteorological station located approximately 2 kilometers from the center of the model domain. The 
wind pattern shows a diurnal cycle that can be explained by the anabatic and katabatic winds of a nearby 
valley. 
In order to determine the particle immission size spectrum which governs the gravitational settling, s 
dust collection experiment across the Dalton road was performed. Along two transects oriented towards the 
prevailing wind direction (nearly north-south oriented), dust samplers were distributed at 5 different distances 
(10, 30, 70, 150 and 300 m) on either side of the road. The collected dust grains were counted by means of 
an image processing system with respect to logarithmically equal size classes. The results show that the 
predominant fraction of the dust has a particle radius of less than 30 pm, peaking at about 7 pm. 
On the basis of the above mentioned modeling principles and the parameters determined during the 
experiment, the pattern of the total dust load was computed. The dust load distribution will be shown in 
Sec. 4 (Fig. 4) and discussed together with the satellite-based distribution. 
3 DUST LOAD ESTIMATION USING MULTISPECTRAL DA 
TA OF THE HIGH RESOLUTION SATELLITE SPOT 
3.1 SPOT imagery 
The polar orbiting satellites SPOT are equiped with two High Resolution Visible (HRV) sensors, HRV1 and 
HRV2. These instruments are capable of operating in the multispectral mode and are sensitive in the three 
bands XS1 (gTeen, 0.50 — 0.59pm), XS2 (red, 0.61 — 0.68 pm) and XS3 (near-infrared, 0.79 — 0.89 pm). 
For broad band applications, they may be switched to the panchromatic mode P (0.51 — 0.73 pm). The 
total swath width is 117 km and the instantaneous field of view (IFOV) at nadir amounts to 20 m for the 
multispectral bands and 10 m for the panchromatic channel. These resolutions match sufficiently well with 
the size of the model grid element of 40* 40 m 2 reported in Sec. 2. SPOT images are acquired only upon 
request, hence there is little chance that data for any given area at a given date is available. Indeed, there 
was no data covering the model domain in summer 1991. The scene which is best suited for our purposes is 
the multispectral image acquired on June 21, 1987, 22:33 UTC (12:33 standard time for the local time zone). 
In Fig. 1 the bands XS1 and XS3 bands are shown. The Dalton Highway and the l^ans Alaska Pipeline are 
clearly visible as bright (in XS1) or dark lines (in XS3). Low reflective areas are identified as rivers or lakes, 
the largest being Toolik Lake.