7th Workshop on Forest Fire Management - EARSeL, European ...
7th Workshop on Forest Fire Management - EARSeL, European ...
7th Workshop on Forest Fire Management - EARSeL, European ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
102<br />
I - PRE-FIRE PLANNING AND MANAGEMENT<br />
The analysis has been performed for some test areas in the south of Italy.<br />
Using two different approaches: (I) classificati<strong>on</strong> performed using a single<br />
date images, and (ii) classificati<strong>on</strong> performed using multidate images.<br />
In the case of single date data processing, we observed that as a whole,<br />
SMA results obtained from Quickbird (overall accuracy 77%) and Aster<br />
(overall accuracy 82%) did not show any significant improvements, whereas<br />
results from TM and MODIS have shown that the use of unmixing technique<br />
allows us to improve at around 7% and 12% the accuracy level for<br />
TM (k coefficient from 57% to 64%) and MODIS (k coefficient from 67% to<br />
79%) respectively compared to the MLC and KNN. These results c<strong>on</strong>firmed<br />
the effectiveness of SMA in handling spectral mixture problems, especially<br />
in fragmented ecosystems as those c<strong>on</strong>sidered for our analysis.<br />
Results from KNN showed improvements around 3% and 5% for Quickbird<br />
and Aster, whereas no significant improvements have been observed for TM<br />
and MODIS data set.<br />
In the case of multidate processing, improvements around 4% were<br />
obtained for the different data set we c<strong>on</strong>sidered.<br />
2 - Final remarks<br />
In this paper, imagery from Quickbird, Advanced Spaceborne Thermal<br />
Emissi<strong>on</strong> and Reflecti<strong>on</strong> Radiometer (ASTER), Landsat Temathic Mapper, and<br />
MODIS, have been evaluated in terms of accuracy and utility for mapping<br />
fuel type and load. Different supervised classificati<strong>on</strong> techniques were compared.<br />
Results showed that each data set processed using different classificati<strong>on</strong><br />
provided satisfactory accuracy levels. In particular, KNN well performed<br />
for Quickbird (81%) and Aster (87%), whereas SMA provided the<br />
best results for TM (64%) and MODIS (79%). Additi<strong>on</strong>al improvement can<br />
be achieved using data fusi<strong>on</strong> approach to merge the spatial and spectral<br />
characteristics of different satellite data set.<br />
References<br />
Lasap<strong>on</strong>ara R., and Lanorte A., 2007. On the capability of satellite VHR<br />
QuickBird data for fuel type characterizati<strong>on</strong> in fragmented landscape.<br />
Ecological Modelling (ECOMOD845R1).