摘要
[目的]分析内蒙古大兴安岭林火发生概率及驱动因素在1987年森林大火前后的差异,了解重大历史事件对林火防控的影响,为未来重大事件发生背景下的林火管理与防火资源合理配置优化提供理论支撑。[方法]基于1980-2018年(39年)内蒙古大兴安岭历史火灾数据,以对我国森林火灾防控有重大影响的1987年森林大火为分界线,采用逻辑斯蒂回归(LR)和增强回归树(BRT)2种林火预测模型对1987年森林大火发生前、发生后和所有年份3个时期的林火发生概率和驱动因素进行分析和比较,同时计算各旗(县)低、中、高3个火险等级的面积,判读不同时期的林火驱动因素和火险变化差异。[结果]1)无论用所有年份还是用1987年前和1987年后的数据建模,BRT模型预测精度均高于LR模型,虽然LR模型预测精度略低,但也可以满足预测需求;2)2种林火预测模型对3个时期的预测准确率均表现为所有年份>1987年后>1987年前,说明在样本数据量足够的情况下,以1987年森林大火为分界线并未提高预测精度,利用所有年份数据建模具有较高可信度;3)气候因素在不同时期均为影响林火发生的主导因素,尤其要关注火灾发生前一年秋季防火期的相关气象指标(平均/最高气温、平均/最高地表温度、日照时数);4)3个时期的火灾中高风险区发生明显变化,内蒙古大兴安岭东部(鄂伦春自治旗东南部、莫力达瓦达斡尔族自治旗大部及阿荣旗中部)3个时期均有较高火灾风险,北部原始林区(额尔古纳市北部)1987年前中高风险区很少,1987年后中高风险区显著增多。导致火灾发生的原因1987年前主要是人为因素,1987年后有关政策和法规的制定使人为因素引发的森林火灾减少,但雷击火次数有所增加。[结论]1987年森林大火重大历史事件的发生使我国森林防火政策发生巨大变化,影响内蒙古大兴安岭林火发生概率及驱动因素的主导�
[Objective]It is clear whether China’s“forest fire in 1987”,a major historical event,has caused changes in the probability and driving factors of forest fires in Inner Mongolia’s Daxing’an Mountains.[Method]Based on historical fire data(1980-2018),taking the“forest fire in 1987”in China as the dividing line,two models,Logistic regression(LR)and boosted regression tree(BRT),were used to analyze and compare the occurrence probability and driving factors of forest fires in Daxing’an Mountains in Inner Mongolia before,after,and throughout the“forest fire in 1987”.Meanwhile,the areas of low,medium,and high fire risk grades in each banner and county were calculated,and the differences in forest fire driving factors and fire risk changes in different periods were interpreted.[Result]1)Whether using all years or data before and after 1987 for modeling,both BRT and LR models demonstrated higher forecasting accuracy compared to each other.Although the accuracy of LR model is slightly lower,it can still meet the prediction requirements.2)The accuracy of both models in predicting at three different periods was observed to be:all years>post-1987>before-1987,indicating that with sufficient sample data,the prediction models based on major historical events such as the forest fire in 1987 did not improve the forecasting accuracy.This finding suggests that the model built using all years of data has a higher degree of reliability in terms of accuracy.3)Climate factors have been the dominant factors influencing forest fires throughout all periods,contrary to previous studies.This study found that it is particularly important to pay attention to relevant meteorological indicators such as average/maximum temperature,average/maximum surface temperature,and sunshine duration during the pre-fire season one year prior to the occurrence of a fire.4)Obvious changes have occurred in the medium-and high-risk areas of forest fires in three different periods.The eastern part of Daxing’an Mountains in Inner Mongolia(the s
作者
周庆
张恒
赵鹏武
周勇
章林
弥宏卓
王嘉夫
赵梦玉
杨泽华
Zhou Qing;Zhang Heng;Zhao Pengwu;Zhou Yong;Zhang Lin;Mi Hongzhuo;Wang Jiafu;Zhao Mengyu;Yang Zehua(College of Forestry,Inner Mongolia Agricultural University,Hohhot 010019;National Field Scientific Observation and Research Station of Greater Khingan Forest Ecosystem,Genhe 022350;Jilin Provincial Academy of Forestry Science(Jilin Forestry Biological Pest Control Central Station),Changchun 130033;Inner Mongolia Forestry and Grassland Monitoring and Planning Institute,Hohhot 010020;Daxing'an Mountains Forestry Science and Technology Institute,Yakeshi 022150;Hohhot Meteorological Service of Inner Mongolia Autonomous Region,Hohhot 010020)
出处
《林业科学》
EI
CAS
CSCD
北大核心
2024年第7期81-94,共14页
Scientia Silvae Sinicae
基金
国家自然科学基金项目(32060344)
内蒙古自治区高等学校青年科技英才支持计划项目(NJYT24042)
内蒙古自治区科技计划项目(2022YFSH0077)
吉林省科技发展计划项目(YDZJ202101ZYTS112)。