Determination Severity of Burn Using Neural Network and LANDSAT TM

An automatic burned area mapping system with Landsat TM has been developed. Given only the boundary of the burned area, the system can determine the severity of burns. Burn severity map shows the texture of burned area, which reflect the affect of topography and biomass status. The SupervisedART-II neural network has been employed. Bands 3,4 and 5 have been used. Post-fire, as well as multi-temporal images have been incorporated in to the network. Using different-sized trainings sets and different dynamic parameters have tested system performance.