MEXICO
BAY
stems
plack-and-white images were enlarged to provide
better discrimination of the lahar-affected areas.
Then a multitemporal color composition was
generated to obtain changes between the two dates of
acquisition. This composite had one image displayed
in blue and green and the second image displayed in
red. Keys to interpret the black-and-white images
and multi-temporal color compositions were
established.
In the eastern portion of the study area,
multitemporal color composites were generated using
three successive acquisitions at 35-day intervals. Two
sets of color composites were created. The first set
used the three images each displayed in a different
color. The second set used two images only with the
first image displayed in blue and green and the
second image displayed in red. The second set was
used in the image interpretation.
To demonstrate the capability of ERS-1 SAR
data in extracting lahar boundaries, a test area along
the Bucao River basin located on the northwest
portion of the study area was chosen. The two-date
acquisitions over this area provided one image at the
end of the dry season when no heavy rainfall was
reported and another image taken at the height of the
typhoon season and rain-induced active lahar was
recorded. The two images were initially filtered to
reduce speckle and preserve edges, then a ratio image
was obtained using the filtered images. This ratioed
image was smoothed using an iterative 3x3 filter,
developed for this project bv J.F. Parrot and called
"Hongkong filter." Then using a threshold, an
automatic segmentation of the image into percentages
of the number of pixels is performed. The segmented
images which now consist of edge-connected pixels
are again smoothed using an iterative filter to obtain
the lahar limits.
For the synergy and complementarity study
of the data sets acquired by different sensors with
different resolutions and on different imaging dates,
the ERS-1 SAR and Landsat TM images were
rectified and properly registered to provide an
effective integration, display and analysis (Taylor,
1993). To minimize positional errors due to terrain
relief, the synergy study was confined to a subset of
the ERS-1 SAR data sets which contain relatively flat
area located in the floodplain. An image-to-map
rectification procedure was used to initially correct
the Landsat TM data. This image was registered to a
1:50.000-scale topographic map (UTM projection)
with 23 ground control points. Using a bilinear
(second-order polynomial) mapping function, a root-
Mean-square-error (rmse) of 0.5 pixel was attained.
441
Nearest-neighbor resampling was performed on the
image to retain the original pixel brightness values.
Then, through an image-to-image procedure, the
ERS-1 radar digital data sets covering the southwest
portion of Zambales were rectified using the corrected
Landsat TM data. Fifteen ground control points were
used in a bilinear mapping function and an rmse of
0.28 pixel was attained.
7.0 RESULTS
7.1 Visual interpretation of black-and-white
SAR imagery for the western portion.
The first image was acquired on 9 July 1993,
at the end of the dry season. No heavy rainfall or
typhoon was reported in the area one month before
the image was taken. The major lahar sites covered
by this SAR image were the Bucao and Santo Tomas-
Marella Rivers in Zambales located on the western
portion of the study area. The downstream portions
near the mouths of the Bucao and Santo Tomas-
Marella Rivers display dark surfaces and field
observations show that lahar deposits composed
mainly of fine material formed the downstream
portion of the river valley which continued to be
flooded by smooth mudflows two years after the
eruptions (Figure 3). Smooth and flat surfaces are
consequently responsible for the low backscatter
values. A trace of slightly agitated stream flow
corresponds to the light-toned linear features. The
downstream portions of the rivers indicate more well-
defined drainage patterns.
Going to the upstream portions of the rivers,
there is a progressive transition in tone from dark to
light. The lighter tone corresponds to dry rough lahar
deposits composed of coarser material upstream. The
rough surface explains the higher backscatter values.
The upstream portions do not show clearly-defined
drainage patterns except for the principal thalwegs.
The third river, Maloma, draining Mount Pinatubo on
the west direct into the South China Sea does not
indicate presence of lahar on the date of imaging.
The second image was acquired on 13
August 1993. A comparison of the two images (July
and August) shows that significant changes occurred
in the intervening 35-day period. Three typhoons hit
the area and induced lahar flow. These included
typhoons Luming and Narsing (local code) from July
25 to 30, 1993 and typhoon Pining on 11-12 August
1993 when heavy rainfall occurred in the Mount
Pinatubo area. Based on PHIVOLCS reports of lahar
events on August 13, high-speed lahars flowed down
the two rivers with varying speeds from a minimum
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
