Use of Multitemporal Indexes in the Identification of Forest Fires - A Case Study of Southern Chile

Abstract

When mapping burned areas, the use of indexes such as: NDVI, NDII, SAVI, GEMI and IAQ (BAI)allow the classification of vegetation status, of which the IAQ index has been designed to clearer identification of areas affected by fires. Another alternative is the multitemporal analysis, which detects changes from one image already classified to another, corresponding to the same surface on different dates. The common denominator applies both techniques separately in order to take advantage of the characteristics of a classified image. Therefore, the current study aims to combine multi-temporality and different indexes to integrate both characteristics and facilitate the identification of burned areas. The study area is located in southern Chile with the corresponding dates from January 15 to 30,2017, where fires consumed much of the endemic forestation of the area. The methodology applied has been based on obtaining and correcting Landsat 8 satellite images prior and after the event. Following this, the different indexes have been calculated to apply the change detection. Thereafter, the results have been integrated to make a multitemporal RGB combination of the new bands where, in the red quill corresponds to the indexes before the event, the green quill to the indexes after the event and the blue quill to the multitemporal difference. As a result of the multi-temporal index combinations, the outstanding index is the SAVI-Multitemporal, which allows 93% visual discrimination of the areas affected by the fires, unlike the SAVI, BAI, GEMI, BAI indexes -Multitemporal and GEMI-Multitemporal with a percentage of discrimination of about 73.33%, 66.67%, 80%, 20% and 86.67%, respectively. Finally, we identified that 157,116,173 hectares of forest have been affected by the fires that occurred in Chile in the studied time period.