The logit model was used to predict the number of fire-days in the Whitecourt Forest of Alberta. The database used included fire (1) and no-fire (0) observations for fire season days between April and October for the 1986 through 1990 period. There were 8,009 observations during this period of which 157 were fire observations. Using four variables, we were able to predict 79.0% of the fire-days and 81.5% of the no-fire-days. The model, Zi=-8.5171+7.6590xAREAi+0.7367xDISTRICTi+2.0478xBUIi+3.9563xISIi, failed to predict 37 of the fire-days and produced 29 ''false alarms''. When this model was tested on fire occurrence data from the Whitecourt Forest for 1991 and 1992 fire seasons it was correct 74.1% of the time. The management implications and limitations of this study are also discussed in this paper.

This study models the daily human-caused wildfire occurrence in Whitecourt Forest, Alberta, Canada. Two models were developed, a logit model and a neural network model. Both achieved similar accuracy when applied to daily prediction (74 vs. 76 %). This paper describes the methods used for the models development and discusses the management implications and limitations of the study.