A Landcover Classification Scheme for High Resolution Imagery and It’s Application to Bushfire Protection in Residential Areas

With the increasing degree of global climate change, bushfires are becoming a major threat to human life and property. A risk assessment of bushfires is dependent on the availability of suitable information on the environment and human activities. Most of the spatial information for fire behaviour prediction is time-dependent, so it is both quite difficult and potentially very expensive to maintain and ensure the reliability of such data. A suitable source of wide coverage is high resolution (≤1m) satellite imagery for risk assessment in bushfire areas. Our objective is to exploit high-resolution stereo data such as IKONOS and extract all possible information such as the vertical dimensions of a forest, the crown size and shape of some individual trees, the location of tree and grass areas, the topography of bare land and the location of housing areas. This was achieved by extracting a DSM, DTM and landcover from IKONOS stereo multispectral imagery. Then we employed FARSITE (Fire Area Simulator http://www.farsite.org) to simulate fire behaviour with several different ignition points in order to assess the risk to housing within residential areas which can be detected by our classification scheme. This case study shows how high resolution stereo images can be exploited to cope with a natural disaster by extracting detailed information on both the natural and man-made environment.