Regional Fuel Mapping Using a Knowledge Based System Approach

Abstract

The aim of this work is the development and application of a method for large extent fuel models mapping with a resolution that supports calculations of potential fire behaviour at local level. The idea behind this study is that the patterns underlying fuel spatial distribution can be identified by analysing examples in a complex geo-referenced data set. Once the patterns are known they can be the basis of a classification for the entire database. In the analysis we use a knowledge based system approach based on induction methods through application of two types of neural networks. The knowledge base constructed for the analysis consists of information gathered from digital maps, satellite image processing, and field work. A layer with the location and attributes of a set of 300 fuel sample points is included to be used as examples. The results reveal that the methodology proposed is successful in three out of the four land cover situations analysed.