关键词: 长白山风灾区, 地下火, 燃烧特征, 发生预测, Logistic回归

Abstract: Objective Based on indoor simulation of ignition, the characteristics of smoldering of sub-surface fires in wind-caused disaster area of Changbaishan Mountain were mastered, and a probability model was established to provide a theoretical basis for the prevention and control of sub-surface fires in this area. Method The sub-surface fuel in the Changbai Mountain wind disaster area with different recovery degrees (recovery area, semi recovery area and non restored area) was used as the research object, and different moisture content gradients (0%, 5%, 10% and 15%) of sub-surface fuel were set up. Through underground fire simulation ignition experiments, the characteristics of smoldering temperature and spreading rate of sub-surface fires with different recovery degrees were mastered. A two-factor analysis method was used to determine the influence of different recovery degrees and moisture contents on the smoldering peak temperature and spread rate of sub-surface fires. A probability prediction model of smoldering of sub-surface fires in wind disaster area was established based on logistic regression model by using two factors of moisture content and depth. Result The limit moisture content of smoldering of sub-surface fires in the recovery area and non restored area from wind disaster in Changbai Mountain was 10%, and the limit moisture content of smoldering of sub-surface fires in the semi recovery area was 15%. The smoldering temperature of sub-surface fires with different recovery degrees was lower, and the self-sustaining combustion time increased with the increase of moisture content of fuel, and the extinction time was the shortest under the condition of high moisture content. The smoldering spread rate of sub-surface fires was slow, with the fastest being only 3.25 cm·h^–1. The interaction of different recovery degrees and moisture contents on the smoldering peak temperature of sub-surface fires was significantly different, among which there was a significant difference in the peak temperatures of different recovery degrees between the conditions of 0% and 5% moisture content, and there was a significant difference in the peak temperatures of different moisture contents among the differently restored area. The spread rate of smoldering of sub-surface fires was affected by the recovery degree and moisture content respectively, but there was no significant difference in the impact of their interaction on the spread rate. The established prediction model of smoldering probability of sub-surface fires had good fitting effect and high prediction accuracy (P<0.01，AUC=0.917). Conclusion The smoldering temperature of sub-surface fires in the recovery area and the semi recovery area of Changbai Mountain wind disaster area is relatively high, with the highest temperature of 640.57 ℃ and 602.02 ℃, respectively. The smoldering spread rate of sub-surface fires in the nonrestored area is the fastest （3.25 cm·h^–1）. The probability prediction model of sub-surface fires based on logistic regression has high prediction accuracy.

Key words: Changbai Mountain wind disaster area, sub-surface fire, combustion characteristics, occurrence prediction, Logistic regression