基于改进的王正非模型结合元胞自动机的林火蔓延预测  

Forest fire spread prediction based on improved Wang Zhengfei model combined with cellular automata

作　　者：田玉萍 金成宇 王斌[1] 李明泽[1] TIAN Yuping;JIN Chengyu;WANG Bin;LI Mingze(School of Forestry,Northeast Forestry University,Harbin 150040,Heilongjiang,China)

机构地区：[1]东北林业大学林学院,黑龙江哈尔滨150040

出　　处：《中南林业科技大学学报》2024年第5期14-25,共12页Journal of Central South University of Forestry & Technology

基　　金：国家重点研发计划项目(2020YFC1511603);中央高校基本科研业务费专项资金资助项目(2572022DT03);碳中和专项科学基金项目(HFW220100054)。

摘　　要：【目的】针对当前我国林火蔓延预测仍存在预测精度不高、普适性差等问题,基于王正非速度模型结合多维元胞自动机的林火蔓延预测模型进行改进,研究了不同区域和分辨率下该模型的有效性,增强了该模型对于不同分辨率数据的适应能力,从而更好地对森林火灾蔓延进行预测,并为林区的火灾蔓延预测和管理提供一种科学合理的技术手段。【方法】使用黑龙江省大兴安岭地区2011年10月28日发生的森林火灾(分辨率500 m)与四川省凉山地区2022年3月29日发生的森林火灾(分辨率30 m)作为数据源,提取发生火灾时刻以及蔓延过程的火线。引入归一化植被指数(NDVI)对王正非模型进行改进,并利用反卷积算法对元胞自动机算法进行改进,将改进后模型输出的结果与实际结果进行精度对比,并比较不同分辨率下的模型精度表现。【结果】改进后的模型精度有明显的提升,其中,黑龙江省大兴安岭试验改进后模型的kappa系数提高了0.0297,面积误差率降低了21.33%,火场预测火点170个,平均过火速度0.75 m/min。四川省木里县试验改进后模型的kappa系数提高了0.1165,面积误差率降低了37.08%,火场预测火点1795个,平均过火速度为4.00 m/min。【结论】改进后的林火蔓延预测模型可以更有效地预测火灾蔓延并计算出最可能过火的火点位置,其预测结果具有高度的一致性和准确性,提高了林火蔓延模拟预测的实用性。与原始模型相比,改进后的模型很好地提高了在不同分辨率数据下的预测精度,能为林火预防和管理提供科学依据。【Objective】At present,the prediction of forest fire spread in our country still has problems such as low prediction accuracy and poor universality.In response to these problems,this study improved the forest fire spread prediction model based on Wang Zhengfei velocity model combined with multidimensional cellular automata,the effectiveness of the model under different regions and resolutions was studied,which enhanced the adaptability of the model to data of different resolutions,so as to better predict the spread of forest fires,and provided a scientific and reasonable technical means for the prediction and management of forest fire spread.【Method】Using the forest fires that occurred on October 28,2011 in the Greater Khingan Mountains of Heilongjiang province(resolution 500 m)and the forest fires that occurred in the Liangshan area of Sichuan province on March 29,2022(resolution 30 m)as data sources,the fires were extracted happening time and the front line of the spreading process.Introducing the normalized difference vegetation index(NDVI)to improve Wang Zhengfei model,and using the deconvolution algorithm to improve the cellular automata algorithm,then comparing the accuracy of the output results of the improved model and the model accuracy performance.【Result】The accuracy of the improved model has been significantly improved.Among them,the kappa coefficient of the improved model in the Greater Khingan Mountains experiment in Heilongjiang province had increased by 0.0297,and the area error rate had decreased by 21.33%,the first fire site is predicted to have 170 fire points,and the average fire speed is 0.75 m/min.In Muli county,Sichuan province,the kappa coefficient of the improved model increased by 0.1165,and the area error rate decreased by 37.08%,the second fire site is predicted to have 1795 fire points,and the average fire speed is 4.00 m/min.【Conclusion】The improved forest fire spread prediction model can predict the fire spread more effectively and calculate the location of the most l

关 键 词：林火蔓延 模型改进 王正非模型 元胞自动机 森林管理 

分 类 号：S762.2[农业科学—森林保护学]