Ocena trafności ustalania stopnia zagrożenia pożarowego lasu metodą IBL względem występowania pożarów

The aim of this study was to evaluate the accuracy of the forest fire risk classification in Poland, developed by the Forest Research Institute (FRI), in relation to the actual occurrence of forest fires between 2016 and 2023. For the analysis, data from the State Forests Information System were used, covering 16,369 fires that occurred in forest areas managed by the State Forests, excluding military training areas, where a different forest fire risk rating system is applied. Each fire record included attributes such as date, burnt area, detection method, fire extinguishment time, forest habitat type, and tree stand characteristics. The forest fire risk degree at the time of each fire was used as the main predictive variable. In this study, it was hypothesized that (1) the burnt area differs significantly between forest fire risk degrees; (2) the daily number of forest fires increases significantly with each higher forest fire risk degree; and (3) both the annual number of forest fires and the total burnt area are positively related to the number of days with the third (highest) forest fire risk degree. To test these hypotheses, aggregated daily and annual data were analyzed. The Kruskal–Wallis test was used to compare differences in burnt area between forest fire risk degrees, followed by a Dunn post-hoc test with Bonferroni correction (Table 1). The results showed statistically significant differences in burnt area across all forest fire risk degrees, with higher risk associated with larger burnt areas. However, the distribution of the data was skewed, with most forest fires being small, but a few very large events influencing the overall average (Figure 1). To assess the association between forest fire risk degrees and the daily number of forest fires, a negative binomial regression model was used due to overdispersion in the data. The results showed a strong and statistically significant increase in forest fire numbers with each higher risk degree. Compared to days with no risk (degree 0), the expected number of forest fires was 58% higher at the first degree, 98% higher at the second degree, and 160% higher at the third degree. The results showed a strong and statistically significant increase in forest fire counts with each higher risk degree (Figure 2). In the annual trend analysis, linear regression and Spearman’s rank correlation were used to assess the relationship between the number of days with the third forest fire risk degree and (a) the number of forest fires and (b) the total area burnt. Both associations were strong and statistically significant (Figure 3), suggesting that the forest fire risk classification is not only consistent with individual forest fire characteristics, but also reflects broader seasonal dynamics. The results confirm that the dynamic forest fire risk classification system currently used by the Polish State Forests, based on the FRI method, effectively captures the probability of forest fires and the resulting burnt area. This method has high predictive power and provides valuable information for early warnings, prevention strategies, and the forest fire protection system, especially in the context of ongoing climate change and increasing forest fire risk across Europe.