On the Mapping of Burned Areas and Burn Severity Using Self Organizing Map and Sentinel-2 Data

Abstract

In this letter, we propose an approach based on the use of Sentinel-2 spectral indices and self-organizing map (SOM) to automatically map burned areas and burned severity. These analyses were performed on a test area in Chania, located in Crete, affected by a fire (around 200 ha) that occurred from July 13, 2018 to July 28, 2018. The investigated area is characterized by heterogeneous land cover types made up of natural and agricultural lands. To identify different levels of fire severity without using fixed thresholds, we applied SOM to the three spectral indices normalized difference vegetation index (NDVI), normalized burn ratio (NBR), and burned area index for sentinel (BAIS) used to enhance burned areas. This is a particular critical issue because fixed threshold values are generally not suitable for fragmented landscapes, vegetation types, and geographic regions different from those for which they were devised. To cope with this issue, the methodological approach herein proposed is based on three steps: 1) indices computation; 2) maps of the difference of the three indices computed using the data acquired from prefire and postfire occurrences; and 3) unsupervised classification obtained processing all the difference maps using the SOM. The obtained results were validated using an independent data set, which showed high correlation with satellite-based fire severity.

Utilizando imagens Sentinel-2 e índices espectrais para análise de severidade em áreas queimadas de origem antrópica: um estudo no sudeste da Amazônia

“…A análise espacial da área de estudo, com índices espectrais, mostrou que ocorreram queimadas de alta severidade, o que de acordo com d 'Oliveira et al (2011) pode retardar a regeneração da área, além disso, os autores destacam que o desmatamento e a queima da biomassa em terras agrícolas é uma ameaça as florestas nativas e pousios na Amazônia. Segundo Lasaponara et al (2020) os efeitos que essas queimadas provocam podem afetar não só a área em que o fogo foi ateado, como pode transbordar para as áreas vizinhas, seja pelas partículas de aerossóis ou pelo fogo. Normalmente, essas áreas vizinhas são abandonadas para regenerar naturalmente, enquanto as áreas que foram intencionalmente queimadas são convertidas para agricultura ou pastagem (d'Oliveira et al, 2011).…”

“…Portanto, diante da tendência de expansão das áreas queimadas e da severidade destas, a hipótese levantada nesta pesquisa é que as técnicas de sensoriamento remoto são capazes de detectar áreas atingidas por fogo e avaliar a severidade das queimadas e, com isso, fortalecer a capacidade de monitoramento da região. Hipóteses similares a essa foram testadas em trabalhos recentes tais como de Chuvieco et al (2019), Penha et al (2020) e Lasaponara et al (2020).…”

Multimodal Noisy Segmentation based fragmented burn scars identification in Amazon Rainforest

“…Each modality provides discriminating information about the same geographical region, helping in mapping amidst adverse conditions like spectral confusion (like due to cloud shadowing) & variability in burn scars making distinguishing between vegetation difficult. Majority of work done in this domain involves methods like auto-correlation [6], self-organizing maps [27], linear spectral mixture model [28], SVM [29], random forests [30]. However, no recent works seem to utilise current deep learning methods like CNN or encoder-decoder models like SegNet or UNet, presumably due to lack of labelled data.…”

Utilizing Sentinel-2 Data for Mapping Burned Areas in Banjarbaru Wetlands, South Kalimantan Province

“…Sentinel-2 imagery can map the burned area and the severity of the fire [30]. Sentinel-2 imagery can be used to be effects of fires and allow for decision-making and land fires [31]. rough Sentinel-2 imagery, locations of severe damage can be identified, so handling can be carried out more appropriately by optimizing existing resources to prevent future fires [32].…”