150 
1 2 3 4 5 7 6 
TM - Simulation bands 
Fig. 3 Spectral signatures of forest classes for 
TM - Simulation bands ( TMS ) 
Fig. 5 
12 3 4 5 
bands 1-5 and 7 
Mixed deciduous forest + Poplar 
(1 - 5) Poplar (pole timber) 
(6, 7) Mixed deciduous forest (old timber) 
It was also possible to seperate cultures of douglas 
firs from pole timber. Older stands, mainly owing 
to their rough crown canopies, have lower reflection 
values in the infrared bands. No significant diffe 
rence could be made out in the visible wavelengths 
(Fig. 8). 
A signature comparison, between the forest stands 
in the Rhine plain and the higher area of the Kai- 
serstuhl, showsd that false classification is un 
avoidable unless topographical data is taken into 
account. For example the signatures of south facing 
mixed deciduos stands are almost the same as the 
signatures of cultures and saplings in the Rhine 
plain (Fig. 9 ) . 
Fig . 4 Forest classes in the feature plans of the 
TM - Simulation bands TMS 2 & TMS 4 
between the age 
stween cultures 
re and the in- 
indsat 5/TM scene 
: differentia 
ls of poplar 
nixed deciduos 
le signature 
red (Fig. 5) ; 
liffering mois- 
ives. 
rered the best 
possibilities of differentiation between stands of 
young pine and douglas firs. Douglas firs had a much 
higher level of reflection in this wavelength than 
pines. However in the visible bands it is the oppo 
site, but the difference of reflection is much less 
(Fig. 6). 
In the summer scene it was easy to identify stands 
of pine from mixed deciduos stands. A satisfactory 
seperation of douglas firs from stands of poplar as 
well as frcm old mixed deciduos stands was not possi 
ble just by using the information from the summer 
scene (Fig. 7). However in comparison the analysis 
of the April scene offered clear signature differen 
ces. 
Ihe summer data set was quite sufficient for a reli 
able seperation of the following age classes of 
mixed deciduos forest: 
a) Cultures and saplings, 
b) Pole timber, 
c) Young timber and old timber. 
4.3 Confusion matrix 
The training areas were classified for the data set 
of the SPOT- and IM-simulation, and the confusion 
matrices shown in tab. 3 and 4 were calculated. 
On the whole the mistakes of ommission and com 
mission in the forest areas were low. The training 
areas proved to be suitable for the classification, 
mostly with more than 90% correctly recognized. 
For the evaluation of the SPCT-simulation poplar 
stands were classed together with the mixed deciduos 
stands. The seperation of the two classes seemed too 
uncertain just by using the SPOT wavelengths. 
Overlapping between the classes cultures and sap 
lings appeared by the IM-simulation data set. 
Although the band IMS5 (middle infrared) offered no 
new possibilities of differentiation for signature 
analysis, the addition of the band offered a better 
recognition of the classes lying between 1% and a 
maximum over 10%. 
4.4 Classification 
For the classification of the Landsat/IM scenes it 
was possible to take into account the phenological