Extraction of burned forest area in the Greater Hinggan Mountain of China based on Landsat TM data

Forest fire is a dominant disturbance regime in boreal forests. The Greater Hinggan Mountain of China is rich in forest resources, but also in a high incidence of forest fires. Aiming the most serious forest fire since the founding of China which happened on May 6^th, 1987 in this area, based on two scene Landsat-5 TM images, we proposed to extract the burnt area and burned forest area in this study. During the extraction, the fire perimeter as well as the rivers, roads and building area were first extracted and masked out by visual interpretation, then four indexes of NDVI, EVI, VFC and DI were calculated and their optimal thresholds for separating burned and unburned forest area were determined according to their histograms and extraction accuracies. The extraction by EVI threshold segmentation was proved to be the optimal one based on the comparison of overall accuracy (99.78%) and kappa coefficient (0.9946). Finally the extracted burnt area and burned forest area were compared with values from official statistics. The remote sensing based extraction which are more objective and efficient, less labor-consuming and repeatable appeared to be more reliable.