Ruecker, Gernot
trunks when weather is wet. Forest canopy has a smoothing effect on relief — when it is removed during fire
effects on small-scale topography affect the radar beam much stronger leading to an increase in image intensity
variations in these areas (Siegert et al 1999). In large areas of the peat swamp forest, were most of the fire
damage class 3 is located, fire causes vegetation death almost without altering vegetation canopy structure.
Therefore, in many test areas, no change in mean backscatter nor in standard deviation of backscatter could be
detected in the images acquired under moist conditions. This allowed for reliable mapping of fire damage and
burn scars in these peat-swamp areas only under dry conditions. Discrimination of three damage classes based on
mean backscatter is better for images taken under dry weather conditions, while under wet conditions, standard
deviation of backscatter is a better indicator for fire damage. This can be interpreted as a consequence of a
decrease in soil moisture during the drought leading to extremely dry soils when exposed after a fire
(Holdsworth and Uhl 1997)), which in turn leads to a decrease in dielectric constant of the soil and subsequently
in backscatter. During moist weather, in turn, the patchiness of vegetation cover and the interaction of the radar
beam with the moist soil and the exposed underlying relief may need to a considerably higher image texture,
manifested in an increase of standard deviation of backscatter for those areas.
Mapping accuracy for fire scars is generally good, although according to the ground inventory for one
concession, the high error makes discrimination of damage classes not feasible. However, according to the air
survey-study, damage class mapping is possible with an overall accuracy greater than 60%. This discrepancy in
results is readily explained by two facts: Firstly, for the area investigated for the block inventory, only images
from the moist period in July were available. As indicated by results of the backscatter study, these are not suited
for accurate discrimination of damage classes. Secondly, assessment from the ground produces results quite
different than damage assessment from the air. Thus, an observer from the air can not identify damage by low
intensity ground fires that leave tree crowns unaffected. An assessment from the air tends therefore to be more
conservative. This may also explain the better agreement between air survey and radar map. However errors of
omission for the slightest damage class are more than 40% for both surveys, indicating difficulties in dedecting
this damage class correctly. Although the ground survey was more detailed, the dataset produced from the air
survey is to be considered as being more reliable, since the area covered by the air survey was much larger,
covering different vegetation and relief types which influence the radar image properties in a different way
before and after fire impact.
The high mapping accuracy for burn scars allows assessment of fire affected areas using standard ERS-2 satellite
imagery become operational. With a lesser accuracy a fire damage estimation was possible. The total area to be
assumed as fire affected is therefore to be considered much larger than was suggested by other investigations
(Liew et al., 1998, MoFEC, 1999). It also has to be borne in mind that accuracy assessment of the radar map
indicate that the estimate of the burned surface may be too conservative, since ground fires may have slipped
detection.
ACKNOWLEDGEMENTS
Many thanks to Anja Hoffmann (GTZ/IFFM) for help and support during the field surveys and to Alexander
Hinrichs (GTZ/SFMP) for providing the block inventories and to both projects for financing the major part of the
study.
REFERENCES
Dwyer, E., Grégoire, J.-M., Malingreau, J.-P., 1998. A global analysis of vegetation fire using satellite images:
spatial and temporal dynamics. Ambio, 27 (2), pp. 175-181.
ESA (European Space Agency), 1997. SAR Toolbox Algorithmic Specifications, Workpackage 213, European
Space Agency, Frascati, Italy.
French, N. H.; Kasischke, E.S.; Bourgeau-Chavez, L.L. and Harrel, P:A., 1996. Sensitivity of ERS-1 SAR to
variation in soil water in fire disturbed boreal forest ecosystems. International Journal of Remote Sensing 17,
pp. 3037-3053.
Fuller, D.O. and Fulk, M. Comparison of NOAA-AVHRR and DMSP-OLS for operational fire monitoring in
Kalimantan, Indonesia. International Journal of Remote Sensing, 21 (1), pp. 181-189.
Holdsworth, A.R., Uhl, C., 1997. Fire in Amazonian Selectively Logged Rain-Forest and the Potential for Fire
Reduction. Ecological Applications, 7 (2), pp. 713-725.
1292 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B7. Amsterdam 2000.