集合卡曼滤波算法对FARSITE模型林火蔓延预测的修正效果研究

Correction effect of ensemble Kalman filter algorithm on FARSITE prediction

针对火源位置输入偏差导致的FARSITE林火行为模型火线预测不准确的问题,提出了一种基于集合卡曼滤波算法的动态修正方法。利用FARSITE对复杂工况下的林火蔓延过程进行数值模拟,以火线位置为待修正参量,以均方根误差(RMSE)为评价指标,对算法的可行性进行了验证,并研究了算法的集合元素个数,观测数据标准差及同化频率对FARSITE预测偏差的修正效果的影响。结果表明:算法能显著提高FARSITE火线预测精度;逐时同化时:集合元素个数为5时,算法的修正效果并不理想,随着集合元素个数增大,样本误差减小,修正效果得到改善,但增大到30以上时,修正能力的提升就不再明显;观测数据标准差大小与RMSE值呈正相关;给定条件下当同化频率由1 h/次降低至2 h/次,整个模拟时长内的误差仍能得到较好控制,RMSE曲线并不会过快增长。

To address the problem of inaccurate fire perimeter prediction that arises from the input error of fire source position of FARSITE(a forest fire behavior model),a dynamic correction method based on the ensemble Kalman filtering algorithm is developed in this paper.With fire perimeter as the parameter to be corrected and root mean square error as the evaluation index,the feasibility of the algorithm is verified through simulations of the spread of a complex forest fire case by FARSITE.At the same time,the correction effect of the algorithm under different ensemble members,standard deviation of the observations and assimilation frequency are investigated.Results show that the proposed method can significantly improve the prediction accuracy of FARSITE.Under the condition of hourly assimilation,the correction effect of the algorithm is not ideal with 5 ensemble members.As the ensemble members increase,the effect is improved because of the reduction of the sampling error.However,when the ensemble members grow above 30,the improvement of the correction ability is no longer evident.The standard deviation of the observations is positively correlated with the RMSE value.Moreover,when the assimilation is performed every two hours instead of every hourly time step,the prediction deviation within the whole time span can be well controlled,and the RMSE curve will not grow too fast.