Research on fire monitoring of Mount Kilimanjaro based on Sentinel-1 images

Le résumé fourni par la source

In this paper, we performed investigations on the potentiality of the Sentinel-1 data, for the characterization and mapping of burned areas. To this aim, we focused on fire occurred on October 11, 2020, in Southeast side of Mount Kilimanjaro. Both VH and VV polarizations were considered. In addition, VV-VH is calculated. The difference of VV, the difference of VH and the difference of VV-VH were computed between Sentinel-1 data acquired before and after the fire using both single-temporal, time-averaged and multi-temporal scenes (to reduce speckle noise effects). The training samples are selected, and the different images obtained in the three situations are used as new bands to participate in the random forest classification algorithm. The purpose is to extract the fire area of the study area. And Planet data with a spatial resolution of 3 meters is used to produce verification data. The accuracy of the results obtained in the three cases is verified, and it is concluded that the fire area extraction result obtained by the method of multi-temporal data is more accurate, and its overall accuracy, Kappa coefficient, Commission, Omission, Commission rate and Omission rate are: 91.27%, 0.82, 0.09, 0.12, 3.71, 4.94, respectively. And, it is concluded that in terms of selecting data time, using the three-phase data after the fire has the best effect.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.