ASSESSING NDVI BASED PHENOLOGY IN DIFFERENT FIRE SEVERITY IN THE KAMAISHI 2017 FOREST FIRE

抄録

 Since fire severity is a term to define fire intensity effects towards forest ecosystem, it is important to find a method to measure fire severity. Furthermore, it is influenced by dryness conditions such as soil moisture and is expected to increase due to climate change. This study aims to explore the potential of using Normalized Difference Vegetation Index (NDVI) based phenology in detecting different fire severity which can enhance understanding towards dryness conditions in relation to fire severity. For this purpose, fire severity was defined using visual observation of scorch crown height which is used as an indicator for the probability of tree mortality. NDVI from Landsat 8 and post-fire observation of 12 selected trees in different fire severity areas in the 2017 Kamaishi forest fire and three trees from unaffected areas were used to analyse the phenology using cloud-free NDVI images. The time series analysis showed the greatest and least changes in NDVI, immediately after the fire, were detected in trees at high and low fire severity respectively. However, detection of low fire severity areas was difficult using NDVI because of the small difference in NDVI between unaffected and low fire severity areas. In addition, changes in NDVI in later months were difficult to interpret without further field observation which highlights the importance of field observation. While the delineation of fire severity areas in the spatial distribution map of NDVI differencing using pre and post-fire images was inadequate because of limited observation points used in determining the threshold for each fire severity level. These findings suggest NDVI based phenology is sensitive only when scorch crown height is at more than 30 % of its total crown height, thus could be used to detect moderate and high fire severity areas which are defined by the probability of tree mortality.