| Analysis of the effects of disturbances on ecological systems has become popular in recent years. Fire is an important disturbance agent of forest ecosystems and must be included as a component in models of vegetation change. The spatial distribution patterns of fire are crucial elements in forest dynamics.; In this dissertation we study the relationship of the probability of fire occurrence with some biological, topographical and climatic factors. The factors considered in this study were vegetation type, elevation above sea level, slope and rainfall. We assume a logistic regression form for the relationship of the probability of fire with these covariates and our model includes a spatially correlated term. This term may be considered as surrogate of unobserved variables such as fuel loads, moisture and others. The model was fitted to data for fires in the Blue Mountains from 1986 to 1993 using Markov Chain Monte Carlo methods and the Hastings algorithm to obtain samples from the full conditional distributions of the parameters involved. Model selection was done using the posterior deviance as a criterion for comparison among different models. Finally, we extend the model to the space-time domain and compare two methods to produce forecasts of the fire risk for a given location.; The results show that the effects of vegetation, elevation and rainfall on the probability of a lightning-caused fire are statistically significant all the time, whilst the effect of slope is significant only during Spring. The highest probability of fire occurred during the Summer. Among the vegetation types, the the estimates for the probability of fire are highest in areas covered by Engelman Spruce-subalpine fir, Whitebark pine and Lodgepole pine. Since these vegetations have been described as typical of mild fire regimes, our results suggest that teir presence in the Blue Mountains might decrease in the future. The lowest fire probabilities were estimated for pixels covered by agricultural and shrublands. The probability of fire decreases with elevation and precipitation. These results are in agreement with what has been reported for the area by other researchers for the area, and explain in part the observed pattern of fire occurrences in the Blue Mountains during the period of study. |
