compo-
on can
an one
'eneous
st min-
ial con-
Bitter-
IS sand
Aiocene
leposits
ce (Q1)
(glacial
(glacial
cked to
lescribe
various
S tested
the lab.
% in 10
the ho-
lignite
20 and
samples
realised
in 2 mm
because
ponents.
| the lab
ch. The
;e of the
ioothing
s a slid-
spectral
the Ter-
ntrast to
Q2 and
g bands
NIR and
| around
2200 nm. It is remarkable that both Tertiary samples in
the VNIR NIR and SWIR range show the highest reflec-
tance values in contrast to them are the values of Quater-
nary sediments in this range not much higher than the
ones of lignite. The striking feature of the mixture curves
is that three out of the four mixtures with a 10 % content
of lignite mixtures have got no effect in the visible region
(up to ca, 700 nm). The mixtures with the 10 % content
and the unmixed sediments have got nearly identical
graphs. These are the mixtures with Q1 (Figure 3), Q3
and T2. But the mixture of lignite and T1 shows a differ-
ent behaviour. The graphs of all three mixtures are very
close together in the visible range as well as in the whole
investigated spectral range and they do not reflect the
variation of the percentage of the components. The grain
particle size is discussed as a possible cause. The samples
Q1, Q3 and T2 have each higher contents of smaller grain
size. Sample T2 consists as a whole of more coarse-
grained material. That is why the proportion sur-
face/volume is shifted to smaller values for sample T2.
Less fine-grained lignite material is necessary to occupy
the surface of the Tertiary sample. Assuming that this is
due to the affinity of lignite for this mixture the effect of
coating then a more coarse-grained sample, even with a
little content, must shift its reflection stronger in direction
to the wetting components as the measurements have
shown. At the same time it can be explained that the
measurements do not shift more in direction to lignite
when the content of lignite is further increased. The extra
lignite particles cannot further occupy the surface of the
sediment particle and do not interact with the reflecting
light.
The samples Q1, Q2 and T2 with medium particle size do
not have such a strong effect of coating. Because of the
proportion of granularity the particles lie next to each
other and so there is further shift of the spectral curves in
direction of lignite with an increasing lignite content. The
situation of the mixture of both of the Tertiary sediments
T1 and T2 among themselves is difficult to understand
(Fig. 4). All measurements within the visible range are
shifted in direction of the Tertiary sediment T1. The shift
within the VNIR and SWIR region is such strong in di-
rection of the T1 graph that the mixtures appear darker
than both of the initial components.
The samples of mixture of Q3 and T2 fit relatively well
the model of linear mixing. The curves are only slightly
shifted in direction of Q3 in contrast to the linear mixing
model. The reason for that is nearly the same grain parti-
cle size for larger particles up to 0,1 mm. Larger particles
have got less possibilities of multiple scattering. Further-
more the other components do not have a chemical affin-
ity among each other as they do between sediments and
lignite.
The investigated samples lead to the conclusion that the
grain size distribution of the components and the chemi-
cal affinity decide considerably what determines the
spectral behaviour of the mixture. Nearly similar grain
particle sizes cause a possible linear behaviour of the
mixture within the regarded spectral range. Different
grain particle sizes lead to non-linear behaviour which
will be intensified in the case of a chemical affinity by the
effect of coating. The results form a basis to the discus-
sion on the use of non-linear mixing models. It must be
emphasised that the procedure of linear unmixing can
only be used when it has been sufficiently proved that
each spectral endmember is spatial separated and that
there are no interactions between the endmember. This is
the case when there is e.g. nearly the same particle size
without an additional content of lignite.
4.2 Classification of vegetation
The investigations of vegetation contribute the monitoring
of the reclamation activities and natural succession proc-
esses. The dependencies of the structures of vegetation on
pedological and hydrological properties are the main
focus of research. The case study shows how spectral and
spatial high resolution airborne scanner data can be used
for specific ecological relevant questions for open-cast
mining areas. The protected sand-dry lawn areas must be
observed in their whole extension because they are en-
dangered by the flooding of the residual holes and by the
involved ground-water increase. Terrestical mapping
cannot fulfil this assignment and it requires a lot of time,
money and staff.
First of all it was checked whether the vegetation units
derived from geobotanical analyses could also be classi-
fied by CASI data. The following species were identified:
e Gray hairgrass swards
e Lichen and moss rich grey hairgrass swards
e Helichrysium arenarium rich psammophyte grass-
lands
e Legiminous rich psammophyte grasslands
e Shrubbery rich psammophyte grasslands
e Calamagrostis epigejos dominated psammophyte
grasslands
* no vegetation.
A Maximum-Likelihood-algorithm with added parallele-
piped limits was also used for the classification. This
procedure is more specific in contrast to other established
procedures (minimum distance) because of its strongly
different standard deviation of single species. Figure 5
demonstrates result of classification for vegetation units
carried out by CASI data in comparison to classification
results on Landsat-TM data. The CASI data have shown
the above mentioned species for the whole sand-dry lawn
area. The band positions of the spatial mode was favoura-
bly disposed in that case. That is why at the time of the
flight the phonological properties of the vegetation units
could be taken into consideration and leave its mark on
the classification. At first a higher aggregation of classes
has been chosen to classify the Landsat-TM data. Dry
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 73
