3 METHOD ICS AND RESULTS OF THE INTERPRETA 
TION OF SPECTRAL SIGNATURE DATA 
The typical spectral course of the remission 
or vegetative objects and of bare soil is 
well known. Primarily the spectral signa 
ture in the whole spectral band 0.4 - 
2.5 /urn is determined by the architecture 
of tne stock. The more oven and horizontal 
the leaves of the stock are and the higher 
i.a. the remission capacity within this 
whole spectral range. In the spectral band 
0.4 - 0.7 /urn the incident radiation is 
absorbed by the pigments in the leaves 
(e.g. by the chlorophyll absorption at 
0.63 - 0.69 /um) and in the spectral band 
1.3 - 2.5 /urn by the water contained in 
the leaves. The depth of the absorption 
bands first of all depends on the primary 
emission with reference to the architecture 
os well as the chlorophyll and water content. 
The degree of soil coverage of stocks without 
complete coverage has a decisive influence on 
the spectral signature, which then become a 
mixed signature. Especially in the spectral 
band C.4 - 0.7 7 um the strength of the 
ground cover-influence is determined by the 
colour contrast between soil and plant (the 
colour contrast is esp. dependent on the 
pigmentcontent of the plant as well as on 
the humus and water content of the soil), 
and in the spectral band 1.3 - 2,5 /um it 
is determined by the contrast between the 
water content of the plant and that of the 
soil. In the thermal infrared band the 
spectral signature reflects the temperature 
of the stock of plants which results from 
the surrounding air temperature, the 
architecture of the stock of plants, its 
evapotranspiration as well as the wind 
conditions. The evapotranspiration itself 
depends on the size of the perspiring leaf 
area per rn^ soil (LAI), on the species 
and site-specific intensity of the vegetable 
discharge and the évapotranspiration pro 
portion of the soil. 
Figure 3 shows four exmples for concrete 
signatures measured with the radiometer from 
the helicopter. The cases discussed here 
ore disticntly marked in the whole spectral 
course of the signatures. Taking - as in 
the biogeographical surveys - extent and 
capacity of the assimilation apparatus as 
well as the accumulated assimilation pro 
ducts as a measure for the productivity of 
stocks it con altogether be derived that a 
stock is the more productive, the lower the 
indices of the spectral signatures during 
the chlorophyll absorption, of the water 
absorption bands and in the thermal infrared 
are, and the higher the indices the near 
infrared remission plateau shows. This 
criterion - in the following named producti 
vity criteriton - was applied to the spectral 
signature data ascertained along the 
trajectories to assess the stocks situated 
in the measuring alignements according to 
their yield formation. The formulated 
productivity criterion was concretized as 
follows and used for the cleavage of a 
quantity of spectral signature indices into 
two parts representing a more productive and 
a less productive target state: if the spec 
tral signature indices of a target in the 
channels 3,7 and 8 are lower than the 
average signature indices of the target 
quantity to be binarized in these channels 
and if the spectral signature index in 
channel 4 is higher than the average index 
in channel 4 it will be grouped into the 
Figure 3. Examples for concrete signatures 
measured with the radiometer from the 
helicopter 
more vital part. If this condition is not 
fulfilled the target will be grouped into 
the less vital port. This criterion is 
applied hierarchically to the quantity of 
targets as long as there is onl„ one target 
in each hierarchy branch left or until the 
targets left can no longer be divided. As 
a result this procedure leads to an order 
of precedence of targets or groups of tar 
gets in relation to their degree of pro 
ductivity. 
In a first duct those spectral signature 
indices were taken together as a target 
quantity which corresponded to alignement 
parts with beet and grain crops. In a second 
duct the two alignement parts were divided 
into two parts of about equal length. 
Independently they were clustered into 10 
classes with the cluster algorithm KMEANS. 
The cluster averages in all 8 channels as 
well as the distribution of the clusters on 
the trajectories were thereby stated. Those 
clusters corresponding to beet and grain 
crops form the target quantity for the 
application of tine productivity criterion. 
In both cases this leads to an order of 
precedence of targets, which begins with 
the most porductive target (or target 
quantity) on both alignements and ends up 
with the least productive target (or target 
quantity). To qualify these statements the 
average soil measuring indices of test pill 
areas were adjoined to this order of 
precedence in a way that the soil measuring 
indices for minimum and maximum stock quality 
of a test pill part corresponds to the lowest 
respectively highest rank of this test pill 
area. The average was formed when the soil 
measuring data of several test pill areas 
could be adjoined to a rank.