Title: Mapping Fire Scars in Global Boreal
Forests Using Imaging Radar Data
Journal: International Journal of Remote Sensing, 2002, Vol. 23, No.20,
4211-4234
Authors: L.L. Bourgeau-Chavez, E.S. Kasischke, S. Brunzell, J.P. Mudd,
and M. Tukman
Reviewed By: Daniel C. Dunning
Abstract
This is a review of a study that demonstrates the effectiveness of utilizing C-band
Synthetic Aperture Radar (SAR), the European Remote Sensing Satellite (ERS),
and Radarsat to map fire scars in global boreal forests. The study found that fire
scars were mappable due to ecological changes that occurred in the burned area
and detectable increases of soil moisture content. The high moisture content
causes an enhanced backscatter signal to be received by the satellite from the
forest making the burned area distinguishable from the adjacent unburned forest
area. Regional ecological differences were found to affect post fire changes as
well, impacting the appearance of C-band SAR imagery. There were four sites in
Canada and an area in Central Russia used for this study. The fire scar
boundaries were mapped with out prior knowledge of the fire locations, and
Canadian Fire Service (CFS) records were used to verify the map results. It was
determined after comparing the delineated fire scar boundaries with CFS records
that utilitizing C-band SAR imagery has a high potential for mapping fire scars in
boreal forests globally.
Introduction
The purpose of this study was to determine if using C-band SAR imagery would
be feasible to map fire scars in boreal forests globally, and to develop techniques
for mapping and monitoring fire disturbed areas on a global basis with the
following objectives: 1) Determine if fire scars are detectable and mappable in
varying ecological conditions using C-band SAR data, 2) Determine if mapping
fire scars with SAR imagery alone is feasible; and 3) Identify any geophysical,
ecological, or temporal conditions which may affect fire scar detection and area
estimation in ecologically different boreal regions. This is important because it
would provide fire service and land management agencies a valuable tool for
mapping areas destroyed by forest fires and monitoring its recovery. In addition
this research is important for global climate research in areas that have frequent
forest fires because of the gaseous emissions generated during the fire and the
post burn biogenic gas emissions that add to the greenhouse gases currently in
the atmosphere.
Researchers used C-band SAR imagery collected by ERS and Radarsat for data
analysis to determine if fire scars can be mapped using radar technology. They
observed that fire scars were three to six decibels (dB) brighter than the
surrounding unburned forests, making the fire scars easy to delineate. It was
determined that burned areas were detectable because of ecological changes
that occurred in the fire scars post burn. Some of these changes included;
removal of tree canopy, exposure of rough surfaces, and increased ground
moisture. The researchers found that this phenomenon occurred only when the
burned areas were wet such as early spring, late fall, or after rain events. The
enhanced brightness allowed the fire scars to be mapped with a moderate level
of accuracy. The increased ground moisture is due to reduced surface albedo,
melting permafrost layer, and a reduction in evapotranspiration.
The benefit of using C-band SAR for fire scar mapping is its ability to penetrate
ground cover. The ERS SAR sensor is a C-band 5.6 cm wavelength imaging
radar with vertical transmit and receive polarization (C-VV). It has a resolution of
30 m and a footprint 0f 100 km X 100km. Radarsat is also a C-band radar
satellite that has a resolution of 30 m with a footprint of 100 km X 100 km. The
benefit of using C-band SAR for fire scar mapping is its ability to penetrate
ground cover. This research indicated that the C-band data has the potential to
be used in conjunction with the Landsat TM data for high accuracy fire scar
mapping and monitoring. The following images (Figure 1) depicts a before and
after of a fire occurring in Alaska. The imagery reveals that both Radarsat and
Landsat can be used to map burn extent, but cloud obscurity is a problem with
the Landsat TM imagery.
Study Areas
To determine if fire scars can be detected and mapped in ecologically varying
boreal ecozones of Canada three ERS study areas were chosen to capture the
west to east differences. The study areas selected were the Northwest
Territories, Ontario, and Quebec (see Figure 2). To capture ecological variation
in the north-south direction, a sequence of three to eight adjacent north-south
images were obtained from each study swath. The North American boreal forest
extends from New Foundland to Alaska with the northern limit ranging from 68° N
Latitude in the Brooks Range in Alaska to 58° N Latitude at the western edge of
Hudson Bay. The southern limit is less distinct and is dependant on precipitation
and soil moisture. Climate ranges from dry with extreme annual temperature
variations in the west to relatively warmer, wetter, maritime climate of eastern
Canada (see Table 1). Fire is more frequent in the drier regions of western
Canada and Alaska than in the eastern regions. Variations in fire frequency are
tied to variations in climate.
Data and Methods
The researchers developed a procedure for this study to map fire scars. The first
step in the procedure was to acquire the data. Data was ordered over regions of
Canada that were known to have high fire occurrence. The data was then
evaluated and rated using a scale of 1-5, with a rating of 3 needed to be visible.
Next, the best rated images were selected, georeferenced, and mosaicked.
Once this was complete the fire scar boundary areas were digitized using GIS
(see Figures 3, 4, &5).
Results
After delineating the fire scar boundaries for the study areas, the data was
analyzed and compared with CFS data for accuracy. The following tables are
the raw data results for the four study areas (see Tables 2, 3, 4, 5, 6, 7, and 8).
Summary of Conclusions
Fire scars are mappable in boreal ecosystems using C-band SAR imagery. For
Ontario and the NWT study sites, fire detection is feasible with SAR. For NWT,
SAR data indicates more area burned than was mapped by CFS. Data
availability for Quebec was limited and analysis inconclusive. Further evaluation
of SAR data collected over Quebec is necessary for any conclusions to be
drawn. Seasonal variations in fire scar visibility occur globally, with the best
viewing season being either spring or autumn. There is some problem
distinguishing fires form wetlands, utilizing Landsat imagery can assist with
distinguishing wetlands from burned areas. For boreal regions in Russia the
SAR data was limited to only two years over a geographic region, and the areas
were not well mapped by fire service agencies. Manual interpretation was used
for this study, but technology exists to automate the process. An improved
method for fire scar mapping and monitoring might use a combination of SAR
and multi-spectral data. Potential future applications for this technology could be
studying fire scar mapping in deciduous forests, fire scar mapping in chaparral,
fire scar mapping in grasslands, and forest reduction in Amazon rainforests.
References
Mapping Fire Scars in Global Boreal Forests Using Imaging Radar Data,
International Journal of Remote Sensing, 2002, Vol. 23, No.20, 4211-4234.