Abstract: Flying fires are an important way of spreading forest fires. In order to investigate the formation mechanism of flying fire ignition, this paper uses camphor pine pine needles as the material to make a fuel bed and camphor pine twigs, bark and pine cones as the flying fire source, and changes the wind speed, the size of the flying fire source and the compression ratio of the fuel bed to conduct flying fire ignition experiments. The optimal modelling data was selected by SPSS and the ignition probability model was developed based on binary logistic regression analysis. A total of 5 760 experiments were conducted for 1 920 experiments for each of the three fly fire sources, with 336, 304 and 376 ignitions for twigs, bark and pine cones respectively. The ignition number curve for each interaction showed an increasing trend and wind speed played a significant influence in the interaction of the other three experimental conditions. The prediction accuracy for the modeled sample of twigs was 75.6% and the total prediction accuracy was 77.3%; the prediction accuracy for the modeled sample of bark ignition was 77.8% and the total prediction accuracy was 83.6%; and the prediction accuracy for the modeled sample of pine cones ignition was 81.2% and the total prediction accuracy was 85.4%. The total prediction rates of the three fire ignition sources were 77.3% to 85.4%, all with high accuracy, which can provide theoretical reference for forest fire prevention.

Key words: Pinus sylvestris var. mongholica Litv, flying fire, probability model, Logistic