1 Spatial Modelling of the Terrestrial Environment - Georeferencial

190 Spatial Modelling of the Terrestrial Environment
section 9.3). Many field-based estimates are also based on head fires, with much lower intensities
reported for flanking and backing fires (e.g. Roberts et al., 1988). Analysis of the
raw data provided by Roberts et al. for a series of experimental grassland fires shows that
the fireline intensity (average ± 1SE; KW m −1 ) of backfires (148.1 ± 23.2; n = 22) was
only 6% that of headfires (2306.9 ± 405.0; n = 39). Finally, the extent to which heat may
be lost to convective and conductive transfer(s) is somewhat unclear, with Ferguson et al.
(2000) quoting convection as providing 40–80% of the total heat loss process (McCarter
and Broido, 1975), a range in broad agreement with the statement that only around one-third
of the evolved heat may be radiated (Vines, 1981). Clearly these effects may explain some
of the difference between the modelled fireline intensities, and the values obtained from
remote sensing. However, there may be effects operating that are currently not accounted
for, and indeed it may be that a much more complex comparison is needed between the
measured and modelled parameters. Further investigations continue on this issue.
9.4.2 Application to Geostationary Satellite Imagery
Despite the current level of uncertainty in the relationship between the FRE and the total
energy liberated during combustion, the strong agreement between the BIRD- and MODISderived
FRE measurements provides confidence in the transportability of the approach
between sensors. One limitation, however, is that neither of these polar-orbiting sensors
can provide very high frequency observations. The experimental BIRD-HSRS sensor can
at best supply only one daily FRE measurement of a fire due to its limited swath width
and orbital configuration, and this can continue for perhaps four or five consecutive days
before the instrument moves out of observing range. This situation is somewhat improved
with the operational MODIS imager where, cloud cover permitting, the two spaceborne
sensors now operating can supply four FRE observations per day of any fire large enough
to be detected. More frequent FRE measurements can only be provided via a geostationary
platform equipped with a suitable detector system having MIR imaging capability. The
current generation of US Geostationary Operational Environmental Satellites (GOES) are
suitable for this purpose, as is the new European Meteosat Second Generation satellite
launched in August 2002. Figure 9.7 provides the first example of remotely sensed FRE
derived from geostationary satellite data. The target was a large forest fire that occurred in
Miller’s Reach, Alaska (61 ◦ N, 149 ◦ W) on 3–4 June 1996 (see Hufford et al., 2000, for
maps of the activity). The Miller’s Reach fire is reported to have burned 1335 ha of forest
on the evening of 3 June and the results in Figure 9.7 were derived by adapting equation
(6) for use with the GOES imager 3.9 µm channel and applying this to the GOES data
presented by Hufford et al. (1999). The values of FRE are of a similar magnitude to the
largest obtained for the Australian fires by BIRD and MODIS, which appears reasonable if
we recall that GOES pixels at this latitude have a spatial dimension of around 8 km, which
is a large proportion of the total length of many of the fire fronts seen in Figure 9.4. Total
FRE release for Fire Pixel 1 as measured by GOES is 8.1 × 10 11 J, and for Fire Pixel 2
is 2.8 × 10 12 J, which as a rough guide equates to 0.7 − 2.2× 10 5 kg and 2.5 − 7.5× 10 5
kg of vegetation combusted (in 3 1 / 2 and 5 hours, respectively) if we assume Burgan and
Rothermel’s (1984) constant heat yield value and the assumption that between 20 and 60%
of the heat is liberated in the form of radiation (see Ferguson et al., 2000). However, as
already stated, the relative importance of radiation and convection may vary from fire to