Evaluation of fire danger rating systems in moderate fire regimes

The establishment of thresholds and the evaluation of fire danger rating systems as a whole are usually based on fire occurrence data. Calculated indices for fire and non-fire days can be compared using a range of techniques, such as percentile analysis, logistic regression and ranked percentile comparison. However, major problems occur in areas with a moderate social and meteorological fire danger. In this case, there are usually only few days with a high fire danger and the probability of fire occurrence on a high fire danger day is also quite low. Traditional evaluation methods can therefore produce misleading results. Yet an evaluation of fire danger rating systems in those moderate fire regimes is still needed to ensure that the best locally adapted index is used for management and mitigation purposes. Alternative techniques therefore have to be based on estimators for fire danger which do not depend on fire occurrence data.

In this paper, a case study is performed for the region of southern Germany (Bavaria). This area is characterized by a temperate climate with occasional dry spells. Forest fires usually occur during such dry periods from March to October. Although the number of fires is relatively low (mean of 33 fires per year on 2.5 million hectares of forested land), fire danger rating systems have a high operational importance. They are used for public awareness raising in times of high danger and help with the scheduling of expensive aerial detection flights. Traditional (fire occurrence-based) methods for fire danger rating system evaluation are presented along with their shortcomings in our area. As alternative fire danger estimators, we examine expert knowledge-based alert levels and dead fine fuel moisture measurements. Their relations to fire danger rating system outputs and potential methods for evaluating fire danger rating systems based on these data are presented. Differences and individual advantages and disadvantages of all of these techniques are discussed. The fire danger rating systems under consideration include both local (e.g. Baumgartner, M-68) and international (e.g. components of the Canadian Forest Fire Danger Rating System, Nesterov, McArthur) indices.