1 Spatial Modelling of the Terrestrial Environment - Georeferencial

188 Spatial Modelling of the Terrestrial Environment
Figure 9.5 Comparison of fire radiative energies derived via equations (6) and (7) applied
to the MODIS and BIRD-HSRS MIR imagery shown in Figure 9.4. Numbers correspond to
those for the individual fire fronts shown in Figure 9.4 and the 1:1 line of agreement is also
shown. Prior to FRE derivation the MIR spectral radiances were corrected for the atmospheric
absorption using MODTRAN parameterized with radiosonde data from Sydney Airport, taken
within 3.5 hr of the satellite overpass
the large wildfires that occurred close to Sydney, Australia in early 2002. Adapting equation
(6) for use with BIRD-HSRS gives the following equation:
FRE MIR = 5.93 × 10 5 (L MIR − L MIR,bg ) (7)
Applying equations (6) and (7) to the corresponding fire pixels making up the fire fronts
visible in the MODIS and BIRD imagery provides evidence of a strong relationship between
the FRE estimates made via the two sensors (Figure 9.5). The agreement appears excellent
if you consider the widely varying spatial resolution, the different spectral coverage of
the sensor band passes and the slight time difference in the FRE measurements. For fires
2, 3 and 10, the underestimation of FRE by MODIS when compared to BIRD-HSRS is
primarily due to the failure of the MODIS fire detection algorithm to identify all ‘fire pixels’
at those locations (see Figure 9.4). The MODIS algorithm was specified prior to launch
and, though it works well on larger events, it is now undergoing post-launch modification
to improve its sensitivity to smaller fires.
In addition to FRE determination, the higher spatial resolution of BIRD allows the fire
front length and the mean radiative energy release per unit length of the fire front to be
determined from the HSRS observations, the latter being a direct measure with the same
units (kW m −1 )asfireline intensity, though in this case referring to the radiative component