Serviços Personalizados
Journal
Artigo
Indicadores
Citado por SciELO
Acessos
Links relacionados
Similares em SciELO
Compartilhar
Investigaciones geográficas
versão On-line ISSN 2448-7279versão impressa ISSN 0188-4611
Resumo
FLORES-RODRIGUEZ, Ana Graciela et al. Comparative analysis of spectral indices to locate and size levels of severity of forest fires. Invest. Geog [online]. 2021, n.106, e60396. Epub 06-Jun-2022. ISSN 2448-7279. https://doi.org/10.14350/rig.60396.
The effects of forest fires on ecosystems are variable depending on the severity of the fire. However, its evaluation in the field means a significant expenditure of resources, either due to its breadth or inaccessibility in the field. Due to this, alternative strategies have been implemented, such as the use of spectral indices derived from remote sensors. However, there is a large number and diversity of these, so in this work a comparative analysis was made in relation to the detection and classification of the severity of a forest fire, which occurred in 2018 in a pine-oak forest. The indices were derived from Landsat 8 images (OLI) and were grouped as: a) Monotemporal (They consider a single image date): NIR = Band 5 near infrared; NDVI = Normalized Difference Vegetation Index; NDWI = Normalized Difference Water Index; NBR = Normalized burn ratio; EVI = Improved Vegetation Index; NBRT = Normalized burn ratio with thermal band BAI = Burned area index; OSAVI = Optimized soil adjusted vegetation index; GCI = Green Chlorophyll Index; SIPI = Structure insensitive pigmentation index; GNDVI =, Green Normalized Difference Vegetation Index; GRMI = Global Environmental Monitoring Index; and b) Bitemporal (For its estimation, two image dates are used, one before and the other after the fire): RdNBR = Relative difference of normalized burn ratio; dNBR = Normalized Burning Ratio Difference; RBR = Relative Combustion Ratio; RI = Regeneration Index; NRI = Normalized Regeneration Index; dNDVI = Normalized difference vegetation index difference. These indices were applied to immediate images prior to the fire and to images obtained during a rainy season after the occurrence of the fire, and were compared with the severity observed in the field during a rainy season after the occurrence of the fire. Subsequently, a severity classification was applied into three categories: unburned, moderate severity and very high severity. This by dividing by natural breaks the values of each pixel in the image. Obtaining thematic maps with the visualization of the severity of the fire for each applied index. Subsequently, for the selection of the index that best fits what was evaluated in the field, a confusion matrix was made, obtaining the global precision and the kappa coefficient. Since a difference was observed depending on the time elapsed after the occurrence of the fire (immediately after the fire and a rainy season later), the PK index (kappa coefficient / global precision) was defined. In this way, the best index for the detection and classification of the severity of the fire was the NBR in general. However, it was observed that the precision is related to the temporality, in this way the best indices defined after the fire were: NBR, GNDVI, RdNBR, Dnbr, RBR, RI and BAI; while those defined after the rains were: NBR, NBRT, NDVI and NDWI. With this, it can be concluded that, although the NBR index turned out to be the best to define the spatial distribution of severity levels of a forest fire in the study area, this does not imply that the same thing happens in all cases, so it is you must select the index that best suits each particular condition. Likewise, the methodological process of this study may support future research where the delimitation of the area in which the classification of the severity of forest fires is going to be made is considered in order to eliminate variability in the range of values generated by spectral indices. In addition to evaluating the largest number of reference sites in the field, as well as ensuring a good distribution of these throughout the study area in different forest fires and defining the largest number of severity classes in the field. In this way, the precision of the indexes will be expanded in order to be able to use this methodology for operational and decision-making purposes that guide the actions that will be applied after the occurrence of a forest fire.
Palavras-chave : burning scar; fire effect; environmental impact; satellite images; Landsat.