Ruecker, Gernot
3.2 Mapped area
The total area mapped as affected by fire in East Kalimantan was 5.2 Mio. ha. The colours in Figure 4 A indicate
the four damage classes: A total of 34% have been assigned the two most severe damage classes 2 and 3.
Although both of these classes indicate that more than 80% of the vegetation have been damaged it is important
to be aware of the fact that the class 3 occurred mainly in ecologically important peat swamp and wetland areas
while class three was typically to be encountered in plantation areas and degraded grasslands. 42% of the burned
area have been assigned damage class 2 (50-80%) 24% have been assigned damage class 1 (25-50%). These
classes typically occured in dipterocarp forest. The burned area extends across the central Mahakam basin and
the coastland towards the slopes of the mountains in the north and west , where the fire extinguished.
3.3 Assessment of accuracy
Assessment of accuracy yielded quite different results for the air surveys and the block ground inventories. For
the accuracy of burn scar detection, the error of omission (burned mapped as unburned) was 5.5% and error of
commission (unburned mapped as burned) was 0.7%. Overall accuracy for discrimination of damage classes was
66.4%. More than 90% of all errors are assignations of an area to a neighbouring class and are therefore
considered to be slight.
For the block ground inventories results are considerably worse. Error of omission for burn scar mapping was
21%, error of commission 1.5%. The overall accuracy for class assignment was only 27.3% indicating that it was
not possible to discriminate damage classes.
Figure 4. Mapped area and Ground surveys. A: The burn scar map. Yellow: 25-50 % damage, orange: 50-80%
damage, brown: >80% damage, canopy remaining, red: >80% damage, soil widely exposed. B: GPS-recorded
tracks of ground surveys in 1998 and 1999. The backdrop to both images is the gamma-filtered mosaic of ER-2
SAR images from August 1997.
4 CONCLUSIONS
Ground and aerial evidence suggest that the marked decrease in backscatter can be attributed to the removal of
the vegetation cover and subsequently higher contribution of backscatter from dry soil. After rainfall, the soil
becomes wet and thus has a higher dielectric constant, leading to a higher radar reflectivity (Ulaby et al. 1986).
In Dipterocarp forests, the fire leads to a removal of the leaves, while the majority of the dead trees remain
standing. This results in pattern of high spatial variability because the radar beam is reflected by remaining
canopy ins some places while in others it may penetrate to the forest floor or double bounce from moist dead tree
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000. 1291