以Sentinel-2A多光谱影像为数据源,利用卷积神经网络模型提取的受害树种空间分布和多时相PROSAIL模型叶面积指数反演差值确定的失叶率耦合的虫口密度,定量获取长白山南麓虫害空间格局.结果表明:2018-2020年共7个时相LAI反演整体精度在88%以上;红松的适宜参考时相为2019年6月,预测与实测拟合R2为0.82,其余树种及全样本2018年6月最佳;虫口密度与失叶率耦合采用线性函数,R2为0.755;落叶松遭虫害面积6174 hm2最大,云杉受害面积比65.19%最大.虫害导致失叶率计算采用的参考时相为受灾前一年6月;虫口密度与失叶率呈线性关系;不同树种受灾空间格局不同,常绿树种重度灾害比例普遍高于落叶树种.
Using Sentinel-2A multi-spectral image as the data source,the spatial pattern of pest damage at the southern foot of Changbai Mountain was quantitatively obtained by coupling the insect mouth density using spa-tial distribution of injured tree species extracted using a convolutional neural network model and leaf foliation rate by the difference of the leaf area index reversed by the PROSAIL model at multiple time points.Results show that:the overall accuracy of 7 LAI inversion in 2018-2020 was above 88%;the optimal reference phase of red pine was in June 2019,R2 is 0.82 and other species in June 2018;linear function,R2 is 0.755;larch pest area of 6174 hm2,and spruce damage area ratio of 65.19%.The reference phase of the leaf loss rate is June of the year before the disaster;the relationship between the pest density and the leaf loss rate is linear;the spatial pattern of different tree species is different,and the proportion of evergreen trees is generally higher than that of deciduous tree species.