Do pheromone traps help to reduce new attacks of Ips typographus at the local scale after a sanitary cut?

The spruce bark beetle, Ips typographus, is causing severe economic losses during epidemic phases triggered by droughts and/or windstorms. Sanitation felling and salvage logging are usually the most recommended strategies to limit the damages. However, any additional control method to limit the economic impact of an outbreak would be welcome. In this respect, the efficiency of pheromone trapping is still controversial or poorly documented. In this 2-year study (2020-2021), at the peak of a severe outbreak in Belgium, we quantified the wood volume and presence/absence of new attacks at 126 sites attacked during the previous year and within 100 m from the initial attack. Each site was randomly allocated to one of three treatments: (1) three crosstraps baited with pheromones, (2) one tree-trap baited with pheromones and treated with an insecticide and (3) control sites with no trapping device. The attacked trees of the previous year were all cut and removed before the start of the experiment and newly attacked trees were removed as they were detected. The trapping devices were only active during spring to target overwintering bark beetles that might have escaped the sanitation cuts and to limit the risk of attracting dispersing beetles from outside the patch during the summer. We found a strong decrease of the attacks relative to the previous year in all treatments, including the controls (more than 50% of the control sites had no new attacks). There was no relationship between the new attacks and the attacks of the previous year. In both years, new attacks were more frequent (presence/absence) in sites with crosstraps (95% Confidence Interval [56-84%] of the sites with new attacks) than in sites with a tree-trap (26-57% - p = 0.02) and to a lesser extent than in control sites (32-63%, p = 0.08). In 2020, the attacked volumes were slightly higher in sites with crosstraps (95% Confidence Interval [3.4-14.2 m³]) than in control sites (0.2-3.5 m³, p = 0.04) and no significant difference was found with tree-trap sites (1.1-6.2 m³, p = 0.38). In 2021, there were no significant differences between the volumes attacked in the control sites (1.8-9.4 m³), crosstraps sites (0.9-6.4 m³) and tree-trap sites (0-2.5 m³). Overall, we found no evidence in favor of the efficacy of pheromone trapping during spring to reduce economic damages at the local scale when combined with sanitation felling and during a severe outbreak. The use of baited crosstraps could even be hazardous as it seemed to increase the occurrence of new attacks probably by attracting bark beetles but failing to neutralize them.

Detection of Larch Forest Stress from Jas’s Larch Inchworm (Erannis jacobsoni Djak) Attack Using Hyperspectral Remote Sensing

Detection of forest pest outbreaks can help in controlling outbreaks and provide accurate information for forest management decision-making. Although some needle injuries occur at the beginning of the attack, the appearance of the trees does not change significantly from the condition before the attack. These subtle changes cannot be observed with the naked eye, but usually manifest as small changes in leaf reflectance. Therefore, hyperspectral remote sensing can be used to detect the different stages of pest infection as it offers high-resolution reflectance. Accordingly, this study investigated the response of a larch forest to Jas’s Larch Inchworm (Erannis jacobsoni Djak) and performed the different infection stages detection and identification using ground hyperspectral data and data on the forest biochemical components (chlorophyll content, fresh weight moisture content and dry weight moisture content). A total of 80 sample trees were selected from the test area, covering the following three stages: before attack, early-stage infection and middle- to late-stage infection. Combined with the Findpeaks-SPA function, the response relationship between biochemical components and spectral continuous wavelet coefficients was analyzed. The support vector machine classification algorithm was used for detection infection. The results showed that there was no significant difference in the biochemical composition between healthy and early-stage samples, but the spectral continuous wavelet coefficients could reflect these subtle changes with varying degrees of sensitivity. The continuous wavelet coefficients corresponding to these stresses may have high potential for infection detection. Meanwhile, the highest overall accuracy of the model based on chlorophyll content, fresh weight moisture content and dry weight moisture content were 90.48%, 85.71% and 90.48% respectively, and the Kappa coefficients were 0.85, 0.79 and 0.86 respectively.

Three-Dimensional Convolutional Neural Network Model for Early Detection of Pine Wilt Disease Using UAV-Based Hyperspectral Images

As one of the most devastating disasters to pine forests, pine wilt disease (PWD) has caused tremendous ecological and economic losses in China. An effective way to prevent large-scale PWD outbreaks is to detect and remove the damaged pine trees at the early stage of PWD infection. However, early infected pine trees do not show obvious changes in morphology or color in the visible wavelength range, making early detection of PWD tricky. Unmanned aerial vehicle (UAV)-based hyperspectral imagery (HI) has great potential for early detection of PWD. However, the commonly used methods, such as the two-dimensional convolutional neural network (2D-CNN), fail to simultaneously extract and fully utilize the spatial and spectral information, whereas the three-dimensional convolutional neural network (3D-CNN) is able to collect this information from raw hyperspectral data. In this paper, we applied the residual block to 3D-CNN and constructed a 3D-Res CNN model, the performance of which was then compared with that of 3D-CNN, 2D-CNN, and 2D-Res CNN in identifying PWD-infected pine trees from the hyperspectral images. The 3D-Res CNN model outperformed the other models, achieving an overall accuracy (OA) of 88.11% and an accuracy of 72.86% for detecting early infected pine trees (EIPs). Using only 20% of the training samples, the OA and EIP accuracy of 3D-Res CNN can still achieve 81.06% and 51.97%, which is superior to the state-of-the-art method in the early detection of PWD based on hyperspectral images. Collectively, 3D-Res CNN was more accurate and effective in early detection of PWD. In conclusion, 3D-Res CNN is proposed for early detection of PWD in this paper, making the prediction and control of PWD more accurate and effective. This model can also be applied to detect pine trees damaged by other diseases or insect pests in the forest.