Application of UAV multispectral remote sensing for monitoring Ailanthus altissima damage

Eucryptorrhynchus scrobiculatus (Motschulsky) and Eucryptorrhynchus brandti (Harold) are two serious wood-boring pests specifically inflicting damage on Ailanthus altissima (Mill.) Swingle. They have caused severe damage to A. altissima, a primary afforestation species used in shelterbelts in Ningxia Hui Autonomous Region. However, no efficient method exists to identify the extent of damage to individual A. altissima at the stand scale, which limits effective pest control.

This study targeted A. altissima in Caowan Village, Qingtongxia City, Ningxia, China, to investigate techniques for identifying pest damage degree caused by E. scrobiculatus and E. brandti using Unmanned Aerial Vehicle (UAV) multispectral imagery. The damage to A. altissima was classified into four levels including Healthy, Lightly, Moderately, and Severely, based on the dead twig rate. UAV multispectral imagery combined with machine learning techniques were used to construct a classification model. To compare differences in the same features across damage grades, the Multiple Wilcoxon Rank Sum Test was applied, while Analysis of Variance (ANOVA) was used to identify pest-sensitive features related to the damage caused by E. scrobiculatus and E. brandti. Furthermore, the classification performance of three machine learning models including Random Forest (RF), Support Vector Machines (SVM), and K-Nearest Neighbors (KNN) was evaluated and compared.

The results showed that the overall accuracies of RF, SVM, and KNN models were 0.866, 0.821, and 0.795, respectively, with corresponding Kappa coefficients of 0.819, 0.762, and 0.722. Among these, the RF model performed the best.

These findings confirmed the feasibility and effectiveness of using UAV multispectral monitoring for detecting and assessing the damage caused by E. scrobiculatus and E. brandti.