Modelling initial attack success on forest fires suppressed by air attack in the province of Ontario, Canada

Abstract

Airtankers are often used on initial attack (IA) to suppress unwanted wildland fires quickly and contain them before they grow large. Skimmer airtankers are commonly used in the province of Ontario owing to its abundance of waterbodies. We examined the influence of airtanker use on IA success on fires actioned by air attack in Ontario using historical fire records and developed three statistical models to estimate the probability of IA success using information available at three different times during the fire response process. These models include information available to the fire management agency at the time the fire was reported, when IA began and during the IA suppression operations. Our findings indicate that the situational information about a fire obtained during IA provides better estimates of the probability of IA success, as demonstrated by increases in the predictive accuracy and area under the receiver operating characteristic curve compared with a model that is based only on information available at the time a fire is reported. Our results can inform pre-suppression planning and suppression resource allocation decision-making, particularly on days during which many new fires are expected to be reported.