ater,
. and
n the
obia-
ps and
7,4 Automatic extraction of lahar boundaries
Using software developed by Jean-François
Parrot at the Laboratoire de Geologie-
Geomorphologie Structurale et — Teledetection,
Universite Pierre et Marie Curie in Paris, France, a
test area was selected along the Bucao River on the
northwest portion of the study area to extract the
limits of the lahar-affected portions of the river. With
the first version of the software, the July and August
images are filtered, then ratioed. This ratio image is
smoothed again using a 3x3 filter. A threshold is
then applied to this last image which results in the
automatic segmentation of the image into percentages
of the total number of pixels. An iterative filter is
then applied to the image to smooth the shapes and
enhance the edge-connected pixels and delineate the
limits of the lahar (Figure 7).
75 Mapping ancient and recent lahar
This exercise aimed to outline the lahar flow
limits and discriminate between ancient and active
lahar by means of the color scheme table shown
below:
July Image
Lahar Coarse Thin
tvpe
August Coarse Light Red
Image Thin Green Dark
The above table was prepared for the study
area along the Bucao River using the July and August
1993 images. This exercise assumes that there is no
significant spatial distortion of the two images such
that these can be matched on a pixel-by-pixel basis.
The differences in backscatter values between the two
images were calculated on a pixel by pixel basis. A
large negative difference meant that a few pixels in
the July image (corresponding to dry fine ancient
lahar with low backscatter values) was subtracted by a
large number of pixels in the August image
(corresponding to thin active lahar with high
backscatter value). This meant that thin active lahar
was flowing over dry ancient lahar. This pixel,
indicating active lahar, will appear dark in the
difference image and blue in the multitemporal color
composite.
A large positive difference meant that a large
number of pixels in the July image (corresponding to
dry coarse ancient lahar with high backscatter value)
Was subtracted by a few pixels in the August image
(corresponding to dry fine ancient lahar with low
backscatter value). This meant that coarse ancient
lahar was flowing over thin ancient lahar. This pixel,
indicating ancient lahar, will appear bright in tone in
the difference image and red in the multitemporal
color composite.
A small difference, or a difference close to
zero, meant the subtraction of large numbers of pixels
in both images (dry coarse ancient lahar with high
backscatter values in the July image and thin active
lahar with high backscatter values in the August
image). It could also mean the subtraction of a few
pixels in both images (dry fine ancient lahar with low
backscatter values in both July and August images).
This meant that thin ancient lahar was flowing over
another layer of thin ancient lahar. This pixel,
indicating thin ancient lahar, will appear grayish in
the difference image and dark in the multitemporal
color composite.
7.6 Radarclinometry
Using the backscatter values of a single
ERS-1 SAR image, slope values are determined and
contours are generated using software presently being
refined by Jean-Francois Parrot at the Universite
Pierre et Marie Curie in Paris, France. A contour
map of the crater of Mount Pinatubo derived through
this process has been produced. The results of this
work will be presented in a separate paper.
7.7 Merged ERS-1 and Landsat TM Data Sets
The July 1993 ERS-1 SAR data and the
Landsat TM 1993 bands 432/RGB data were merged.
The image was produced using the surface modelling
routine of the Earth Science Image Processing
Package (ESIPP) developed by G. Taylor and L. Balia
of the University of New South Wales. This
algorithm uses radar data as a pseudo-topographic
surface and the Landsat TM bands apply color to this
surface. This technique allows for true integration of
the two data sets (Taylor, 1993).
The merged image shows vegetation in red
tones, lahar deposits, mined areas and scarce-
vegetation areas (bare soil and exposed rock surfaces)
in various shades of green, water areas in black and
built-up areas in white. The colors emanate from the
Landsat TM, while the strong white tone of the built-
up areas is contributed by the ERS-1 radar
backscatter. Areas of high relief are strongly
enhanced by the radar information; however, the
mountainous portions of the merged image have high
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996