An expert system with uncertain rules based on Dempster-Shafer theory

We study the aggregation of spatial remote sensing data (slope, aspect, soil depth and rock permeability) to assess the risk of desertification of a burned forest by an expert system. Belief functions which are assigned to data classes are propagated through expert rules. Uncertainty in data because of subsampling error and uncertainty in rules are considered during the processes. Experiments show the effectiveness of taking into consideration the uncertainty in data and rules.