The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part B7. Beijing 2008
downloaded from GLCF (Global Land Cover Facility) -
University of Maryland server (Table 2).
Year
Scale
1964
1:23 000
1975
1:29 500
1984
1:25 000
1995
1:26 000
2004
1:26 000
Table 1. Date and scale of aerial photos
Sensor
Spectral
range
Bands
Pixel
resolution
Registration
date
LANDSAT
MSS
0.50-1.10
4
60
31.05.1979
LANDSAT
ETM+
0.45-2.35
6
30
13.06.2000
Table 2. Sensor characteristic
Thematic vector maps (geological, hydrogeological,
geomorphological, water ecosystems, soils, forest and non
forest communities, fauna, flora, water resources and other) and
tabular descriptive database (the MS Access format) concerning
particular elements of the SNP environment refer to each vector
layer. Also topographic maps of the Park area, scale 1:10 000,
were included in the data set.
2.2 Methods
The aerotriangulation procedure was difficult due to poor
quality (both radiometric and geometric) of processed photos
(those dating back to 1964, 1975, 1984). Therefore, modem
methods of digital image processing, such as filtering, image
enhancement, or preliminary colour balancing were applied to
upgrade the quality of those photogrammetric materials, and to
improve the possibilities of their photointerpretation. The
ImageStation Automatic Triangulation photogrammetric
software was used to perform measurements, necessary for
adjustment of the aerotriangulation block. In the first stage,
interior and relative orientations were performed for all photo
blocks. Main difficulties in the relative orientation stage were
related to the identification of Grand Control Points (GCP)
because of specific character of the SNP area (70% of its
surface area covered by active dunes, lakes and the Baltic Sea).
For all aerialtriangulation projects the signa nought was
between 4-6 m
The Digital Elevation Models (DEM) were produced using
DEPHOS (Mapper Stereo) digital photogrammetric workstation,
with 0.9-1.2 m accuracy. DEMs for the whole investigated area
from each epoch were converted from vector format to a grid
with a 3-metre resolution. DEMs derived from aerial
photography provided relief data, which were used to indicate
selected features changes of the SNP landscape.
In the next step orthophotomaps (scale 1:5000 and 0.5 m
ground resolution) were generated from the aerial photos for
each period of time with ORTHO ENGINE, the PCI Geomatica
module.
On the basis of stereoscopic observation of air photos, contour
maps of dunes that occur within the SPN area were generated
for all year group photos. Contours of particular dune forms
were created through digitalisation of dune skeleton lines. To
verify the correctness of vectorized dune forms, curvature maps
were prepared, as well as elevation maps of the 0.5 m isolines.
Layers showing lines and water course lines were generated on
the basis of DTM, with the help of Geomedia Grid function,
which determines the curvature area in a given point along the
slope inclination direction.
The processing of satellite data was done with the use of ENVI
(Environment for Visualizing Image) IDRISI and PCI
Geomatica software.
In order to minimize the impact of atmosphere on the values of
reflection recorded on images, the images were atmospherically
corrected, with the use of ATCOR 2 (Richter, 1996), module of
PCI Geomatica.
Based on the unsupervised classification, colour composite
images and orthophotomaps (derived from aerial photos), the
supervised classification (The Maximum Likelihood) was
performed (Fig.L). Seven classes of the land cover were
defined:
1. Dune 1
2. Dune 2
3. Water
4. Forest complex
5. Meadows and pastures
6. Agricultural
7. Coastal dune forest
Next, Normalized Difference Vegetation Index (NDVI) was
completed for LANDS AT temporal data (Fig. L).
Figure 1. Classification and NDVI results.
3. RESULTS
Multi-temporal aerial photography, DEMs, orthophotomaps,
thematic vector layers and satellite images have been integrated
for the changes detection of the unique environment of the
Slowinski National Park. Full integration of multi-temporal data
made it possible to conduct an analysis in GIS environment and
to compute maps quantifying the features of landscape changes.
All processed images, integrated with thematic vector maps
enabled to get complete information about environmental
components revealed in raster and vector data.
The differential maps, which had been prepared in GIS
environment, made it possible to perform quantitative analysis
of horizontal shift of selected dunes, as well as observation of
changes in land occupied by them in consecutive time intervals
(Fig. 2. and Fig. 3). Based on the observation of the dune’s