are PRESDo
Control points
f data and in
time (scale
round contro)
S. Data are
> resolution of
rage capacity
ution 550 MB
Jument not to
ation error is
xels (10 m),
rectification of
2 coordinates,
1 into Gauss:
the upper left
nto complete
is worked out
order not to
el, ie. real
med without
uded contrast
partly show
inaccuracy of
was cut outin
wo stripes stil
feathering is
overlap area
getation index
jan areas are
f evening and
ing picture"
(3)
d Difference
ulated on the
3and5) (4
emporal Toti
composite al
imposition for
classification
ware reasons
format. So the
ere in floating
signed intege!
n in detail
Band Contents
No.
1-10 [Multispectral bands 1 - 10
11 Calibrated thermal infrared data of the midday flight
12 Calibrated thermal infrared data of the evening flight
13 Calibrated thermal infrared data of the morning flight
14 Difference of temperature values between evening flight
and morning flight
15 Vegetation index NDVI calculated from bands 5 and 7
Table 3: Chosen Bands of the Multispectral and Multitemporal Data
Composition
3.6 Investigation of Band Correlations
Correlations for each band combination are calculated for
the three flight areas and for certain sites. Table 4 gives the
the general results.
Correlation? | Band combination
”> 0,95 1-2,2-3,3-4,3-5,4-5,7-8
095»52»0,90]1-3,2-4,2-5
09»2»0,85 |1-4,1-5,3-6,4-6,4-10,5-6,5-10,6-9,
9-10
0>r2>-0,5 3715,6-15,10- 15, 11 - 15,12-15,13-15
-0,5> 1-15,2-15,4-15,5-15
Table 4: Correlation r? of certain Band Combinations out of the
15-band Total Data Set (description of bands see Tab. 3)
The following conclusions can be made:
The band combinations 1 - 2, 4 - 5 and 7 - 8 highly
correlate. Information on bands 1, 4, and 8 are mostly
included in the bands 2, 5, and 7. It is no coincidence that
the spectral areas covered by these bands are not used by
Landsat-TM, so these bands can be ignored in
multispectral classifications without a serious loss of
information.
Very little correlation values can be registered in any
combinations with band 15, (vegetation index NDVI).
Negative correlation can be explained regarding the
calculation formula of NDVI. Only in combination with
bands 8 and 7 there is a positive correlation of band 15.
4. ANALYSIS OF DATA FOR SOIL SEALING
IN URBAN AREAS
4.1 Registration Problems of Soil Sealing Degrees using
Aerial Photos and Scanner Data
Problems occur if items are covered or hidden (trees,
underground car parks and planted greenery on roofs):
There is no other way for all remote sensing techniques
than looking down to earth's surface. In the case of passive
Optical sensors there is no chance to look through the
Upper surface material, so everything underneath is not
489
recorded. As a consequence soil sealing under leaved
trees can not be detected. Only with visual interpretation of
the surroundings or with additional expert knowledge these
areas can be identified.
Furthermore sealed soil of underground car parks, whose
surfaces are covered with vegetation, or planted greenery
on roofs can not be registered with remote sensing
techniques. This is a specific problem of city centres and
new residential parks where these kinds of surfaces cover
up to 15 % of the total areas.
Precision of soil sealing degree values derived from remote
sensing data:
Remote sensing methods can hardly deliver the same
exact values as a terrestrial investigation.
This includes the geometric precision as well as the
contential precision. Although there are a lot of defined
indices to describe soil sealing the "soil sealing degree" is
the best index reproduced by remote sensing techniques,
because it does not pay much attention to surface material.
Other indices, e. g. the "soil function index", which is more
sensitive to drain capacity can hardly be reproduced when
using remote sensing data. It can only be derived if
correlated with the soil sealing degree. Even with the
highest geometric resolution small pavement areas or joints
are not recordable.
Rectification of Scanner Data:
The rectification of scanner data is a very time consuming
and costly task. A large number of ground control points
has to be found manually (without automation) and the
actual rectification done by a special and expensive
software. The costs of data rectification are estimated 2 - 3
times higher than the costs of data acquisition. Possibly by
means of very high resolution and precise GPS navigation
instruments such high costs could be reduced decisively.
This latest technique enables to rectify automatically
knowing the precise aircraft position during the whole flight.
Such increasing efficiency would make the use of scanner
data much more attractive to answer various scientific
questions.
Panoramic Distortion by Wide Scan Angle:
The scan angle can reach a maximum of 43° declination
from Nadir. This fact causes immense distortions in
peripherical zones of scanner stripes, large shaded areas
and blind spots, as well as the reproduction of vertical
surfaces, e.g. house walls. To reduce this effect the wide
side overlap of 40 % is chosen and only the central area of
a stripe with its small distortion is then taken at the expense
of a more costly acquisition and analysis.
4.2 Vegetation Index as an Instrument to identify Soil
Sealing
The vegetation index gives an very precise overview,
where earth surfaces are covered with vegetation.
For a multistep analysis strategy the vegetation index NDVI
should only be used to separate areas which are covered
by vegetation and therefore unsealed (wrong interpretation
of underground car parks with plants on surfaces and
planted greenery on roofs are inevitable).
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
