ulat- 
vari- 
| was 
, us- 
uced 
ough 
baly 
ands 
r the 
reas 
plain 
with 
pari- 
) the 
10WS 
' the 
(the 
gets 
(see 
y re- 
ents 
] 4, 
lting 
and 
ND 
ve a 
this 
the 
and 
g to 
(5) 
The resulting datalayer is displayed in figure 1 (C) 
with a resolution of 120m*120m. There are only three 
differentiations made in the display of the residual: 
1. grey: within a. 
2. black: below mean — a. Track systems and high 
density built-up areas and part of the forest areas 
(caused by the sun exposed steep slopes). 
3. white: above mean + o. Water, part of the for- 
est areas (caused by the not-sun exposed steep 
slopes). 
Doing the same operation for the high resolution 
dataset (using band 6 original 30m+30m (that means 
always 16 equal pixels) and the above described re- 
sulting synthetic band 6 in 30m»30m) results in a 
much more differentiated residual with typical devia- 
tions for individual classes (example: the runways of 
the airport have typical deviations). This newly cre- 
ated datalayer was added to the original bands and a 
maximum likelihood classification led to excellent re- 
sults within the focused urban study area? (for clas- 
sification and a discussion of its results see Wüthrich, 
1991). Amongst the rural classes, it was possible to 
separate 16 urban classes. 
BRIGHTNESS TEMPERATURE AND UR- 
BAN STRUCTURES 
The thermal pattern of a landscape is strongly influ- 
enced by the surface characteristics, especially in ur- 
ban agglomerations (Oke, 1987). 'The regression anal- 
ysis represents a method to quantify the influences of 
different built-up areas on the spatial distribution of 
radiation temperatures. In this context the following 
topics have to be mentioned: 
e The longwave emission/radiation temperature is 
of fundamental importance for the energy bal- 
ance, especially during autochthone weather con- 
ditions. The net radiation is described in equa- 
tion 6. 
E z(StAD0xü-pgrB£bi-5Et (6 
S direct solar irradiance 
Eg | diffuse solar irradiance 
  
Sfor the classification a further dataset was created: a vari- 
ance filtered vegetation index (NDVI) in order to get the dif- 
ferences of uniform areas and spatially rapidly changing areas 
The accuracy of the result was an average 8096 of correctly 
classified pixels within the training areas. The area within 
Switzerland was superimposed with the communal boundaries 
to get classification results for each community. So these results 
could be compared to conventionally achieved data (such as A: 
planimetered on topographical maps using aerial photographs 
and B: official areastatistiks of Switzerland) 
425 
p Albedo 
Eı | atmospheric counter radiation 
Eı 1 terrestric longwave emission 
E"* net radiation 
The radiation temperature Tg is directly con- 
nected with the longwave emission through the 
law of Stefan-Boltzmann (equation 7) . 
Ei 1— 0TR (7) 
E; 1  terrestric longwave emission 
c Stefan-Boltzmann-constant 
(5.6697 « 1079?Wm-? K-*) 
TR radiation temperature 
e The quantification of the landuse influences im- 
proves the understanding of the energetic pro- 
cesses between surface and atmosphere. 
e This quantification is a prerequisite to simulate 
synthetic thermal images under the conditions 
of a modified landuse (e.g. as a consequence of 
urban planning) 7. 
ANALYSING THE INFLUENCE OF THE 
LANDUSE ON THE RADIATION TEM- 
PERATURE 
Using a multiple linear regression analysis, which ex- 
plains the radiation temperature Tr as a linear com- 
bination of the percentages of the different landuse 
classes i (i = 1 to n classes) within the pixel Pj,,, 
multiplied with a regression coefficient and summed 
up with a regression constant. The spatial resolution 
of the classification data (30m * 30m) and the ther- 
mal data (120m * 120m) differ, so the percentage of 
the landuse classes is calculated from 16 pixel (4«4), 
which build up one thermal pixel. As the satellite's 
sensor measures an energie flux, the values of band 6 
had to be transformed to brightness temperatures us- 
ing the method of Schott and Volchok (1985). The 
regression results for the urban area are listed in ta- 
ble 3. The explained variance reached 88.4% . Using 
these results, it was possible to recalculate the radi- 
ation temperatures inside the study area. The origi- 
nal data received from band 6 is displayed in figure 3 
(A) for a selected frame. The recalculated /simulated 
radiation temperatures are placed on its right side in 
figure 3 (B). Calculating the difference of the original 
data Porig,, Minus the synthetic data Psynt,, for each 
pixel indicates that 67% of the resulting Pa:;;,, are 
within the range of +0.9K (with maxpos = +3.6K 
and mazneg = —5.2K). 
  
"The influence of vegetation changes on the net radiation 
has been quantified (Parlow and Scherer, 1991) for an area in 
Swedish Lappland