Development of the Forest Fire Detection Algorithm for Gcom-c1/sgli

This paper presents a development of the forest fire detection algorithm for upcoming GCOM-C1/SGLI sensor. Several existing fire detection algorithms which are applied to GOES/VAS, NOAA/AHVRR, Terra Aqua/MODIS sensors mainly use the MIR fire channels(3-4μm) for key role of the fire detection algorithm. Meanwhile, because of lack of the MIR fire channel on GCOM-C1/SGLI, fire detection algorithm must be developed using general land/atmosphere observation channels. For this reason, we investigated the significant channels for the fire detection process and determined the frameworks of the fire detection algorithm for SGLI. The proposed fire detection algorithm mainly uses the second principal component score of SWIR 1.6μm and 2.2μm channel’s radiance, and the second principal component score of NIR 0.8μm and SWIR 1.6μm channel’s radiance to detect fire pixels. The algorithm is organized with complex of the various threshold tests. Comparison between the fire detection result on this algorithm and ASTER fire detection result under 30m high resolution data using an automated fire detection algorithm for ASTER(Giglio et al., 2008) shows that the detection accuracy of this algorithm is 14% of producer’s accuracy and 87% of user’s accuracy and fire pixels which has a fire fraction over 2.5% or a fire temperature over 1000[K] can be detected using developed algorithm.