Reconstructing forest fire history -- identifying hazard rate change points using the Bayes' Information Criterion




Reed, William J.

Journal Title

Journal ISSN

Volume Title



Graphical analysis of data from a time-since-fire map often suggests temporal variations in historical forest-fire frequency. In this article it is assumed that changes in the fire hazard rate, if they ever occurred, happened at distinct change points, separating epochs during which the hazard rate was constant. A methodology for determining the most plausible number of change points, using the Bayes' Information Criterion, is developed. It is based on an overdispersed model (with corresponding quasi-likelihood function) for the burning or survival of units of the forest. The method is applied to two datasets from time-since-fire maps in the Canadian Rockies. In both examples a single most plausible model stands out. In each case maximum likelihood estimates of change points and of the hazard rates which prevailed between them are calculated.



BIC, change points, fire hazard rate, model selection, quasi-likelihood