Impact of Different Fire Severity Levels on NDVI Phenology at Different Severity Levels in the 2017 Kamaishi Forest Fire

Fire severity has been associated with ecosystem responses such as erosion, sedimentation production and tree mortality which indicates different variation of fire severity found in a burned area caused various effects on ecosystem both in the forest and whole basin. These variations are based on fire intensity, type of vegetation, topography, weather conditions and others. By using remote sensing data and ground truth from post fire observation, these variations can be observed. Most research had focused on identifying hot spots and burned area with remote sensing and assessing the impacts of forest fire on biodiversity and water quality but few research related on different fire severity levels and NDVI phenology in Japan which has many mountainous terrains. An opportunity presents itself, when a large forest fire broke out in Kamaishi, Iwate Prefecture, Japan on 8 May 2017 with a burnt area of 413 ha which is larger than the total burnt area for the whole of Japan in 2016. In this research, Normalized Difference Vegetation Index (NDVI) from Landsat 8 and post fire observation of selected trees affected by different level of fire severity were used to analyze the NDVI phenology. Two datasets were used, NDVI images which were cloud free and average NDVI which were derived by averaging the NDVI images into two images per month. Next, these NDVI were extracted based on the location of trees observed and time series were created. Time series using NDVI average showed more uniform changes on NDVI phenology than time series using cloud free NDVI images as expected due to smoothing of the images. Results from time series using both datasets indicate the changes in NDVI phenology are more distinct in trees which were located in high and moderate fire severity burned areas but not in low fire severity burned and unburnt areas. This indicates the possibility of underestimation by NDVI based satellite analysis for low fire severity and unburned areas because NDVI is sensitive towards trees crown. However, these findings also suggest NDVI based satellite analysis has the potential to be used in fire severity estimation but further analysis based on more ground truth is needed for better understand of phenology affected by fire.