APPLYING NEURAL-NETWORK TECHNOLOGY TO HUMAN-CAUSED WILDFIRE OCCURRENCE PREDICTION

Human-caused forest fires are a serious problem throughout the world. Believing that there are predictable characteristics common to all fir es, we analyzed the historical human caused fire occurrence data for t he Whitecourt Provincial Forest of Alberta using artificial neural net work and geographic information system (ARC/INFO) technology. These da ta were also analyzed using logistic regression analysis (the binary l egit model), which served as the ''domain expert'' to identify the imp ortant input variables. A 314 fire and no-fire data set for the period 1986-1990 was used for training. The observations were whether at lea st one fire occurred, on a certain day, in one of the eight geographic zones defined within the study area. The models developed were tested using data from the 1991-1992 fire seasons, which had 58 fire observa tions. Using as input variables the Canadian Fire Weather Index for th e day, area in km(2) of the geographic zone, and district (a 0/1 dummy variable from the logistic regression model, which accounts for obser vations within a forest district where human use is higher), the resul tant model had four input nodes and two output nodes, and correctly pr edicted 85 percent of the no-fire observations and 78 percent of the f ire observations.