Forest change detection plays a vital role in remote sensing research, serving as a cornerstone for ecological protection and sustainable environmental management. While existing research on change detection primarily focuses on urban areas and croplands, forest change detection remains underdeveloped. As the demand for real-time monitoring of forest changes grows in response to challenges like climate change and environmental degradation, advancements in remote sensing technologies make it possible to bridge this research gap. The complexity of forest change detection stems from high seasonal and inter-annual variability, with fluctuations often showing strong similarities across years. In this paper, we introduce FCDNet:A Multi-Scale Attention Networkfor Forest Change DetectionUsing Dual-Temporal Very-High-Resolution Remote Sensing Images. The proposed network is designed to enhance the precision and robustness of change detection through the integration of multiscale feature aggregation and an adaptive channel attention strategy. This design enables FCDNet to sensitively identify fine-grained forest alterations while maintaining strong adaptability to spatial and temporal variability in the input imagery. Comprehensive experiments confirm that FCDNet delivers superior performance compared with existing state-of-the-art approaches, showing notable advantages under challenging scenarios characterized by seasonal transitions and inter-annual fluctuations. FCDNet achieves impressive F1 scores of 78.51%, 90.48%, and 92.57% on the Forest Change Detection (FCD), LEVIR-CD, and WHU-CD datasets, respectively, delivering state-of-the-art results. This approach holds significant potential for advancing forest monitoring, providing a reliable tool for researchers and policymakers to better understand and protect forest ecosystems.

Pine wilt disease (PWD) is a destructive pine disease with a fast onset rate, high mortality rate, and high difficulty in prevention and control. Accurate and efficient monitoring is the foundation of disease prevention and control. This study aims to explore the potential of red-edge and blue-edge inflection positions from hyperspectral data in monitoring physiological changes and early detecting PWD. We obtained samples of Japanese pines and measured their needles hyperspectral data (wavelength range: 350-2500nm) and physiological parameter data, and obtained hyperspectral drone images of Chinese red pine (wavelength range: 400-1000nm). We used linear fitting algorithms to investigate the linear relationships between vegetation indices and physiological parameters and tested the sensitivity of vegetation indices for PWD early identification using linear discriminant analysis (LDA). The results showed that the indices of the blue-edge inflection position and the red-edge inflection position can reflect the changes in needle pigment content and moisture content, with the 4 point linear interpolation of the blue-edge point showing the best fit for water content. At the needle scale, linear interpolation indices of the blue-edge inflection position showed high accuracy in identifying both early and full-stage PWD. However, the accuracy of these indices decreases when using drone data. We concluded that the developed blue-edge inflection position vegetation indices can be used for PWD early identification. However, further optimization of band selection is needed to improve their application with drone data.

Pine wilt disease (PWD) is a devastating forest disease and has been listed as a quarantine pest in 52 countries around the world. Early identification of the affected trees and timely removal of them from the forest is crucial to control the spread. This study aims to explore the potential of hyperspectral data on early identification of PWD and exhibit the common spectral features, from early-infected tree crowns and needles, and from different species located in different regions. Two types of hyperspectral data were used and compared. One was using drone-based hyperspectral images with a spectral range of 400 – 1 000 nm and a resolution of 0.11 m. The images were analyzed at the individual-tree level. The other was using hyperspectral reflectance from sampled needles with a spectral range of 350 – 2 500 nm. It was used for the analysis at the needle level. We used linear discriminant analysis (LDA) to quantify the separability of spectral reflectance and first-derivative reflectance from the healthy and early-infected samples. The results showed that the red-edge bands were more sensitive than the other bands at both individual-tree and needle levels, and the first-derivative of red-edge bands achieved the best early recognition of the disease with 0.78, 0.72, and 0.85 accuracy at the individual-tree level for Chinese red pine and at the needle level for Japanese pine and Korean pine. We concluded that red-edge bands were the most informative bands with stable sensitivity at different scales and for different species.

Since the pine wilt nematode spread and was discovered in China in 1982, it has caused severe damage to the pine forest and has caused a substantial finical loss. Detecting infected trees and removing them from the forest as early as possible is crucial to prevent its spread by the insect vector longhorn-beetles. This study aims at developing methods to detect and map infections using hyperspectral drone images. We inventoried 391 pines in middle east China and recorded them as healthy or early-, middle-, late-stage infected trees. The hyperspectral drone images were obtained with 0.11 m resolution and wavelength from 400 to 1000 nm, covering from red band to near-infrared (NIR). We used the successive projections algorithm (SPA) to select the sensitive bands and the support vector machine (SVM) algorithm to classify trees into different health statuses. The classification resulted in high accuracy during the middle and late-stage infection, while separating healthy and the early -stage infection was challenging. The developed method could map Pine wilt nematode infections and guide the sanitation felling as a crucial disease control measure.

Accurate and fine multiple tree species supervised classification based on few-shot learning has attracted close attention from researchers, because the sample collection is often hindered in forests. Prototypical networks (P-Nets), as a simple but efficient few-shot learning method, have significant advantages in forest tree species classification. Nevertheless, the overfitting phenomenon caused by the lack of training samples is still prevalent in few-shot classifiers, which brings challenges to training accurate classification models. In this study, we proposed a novel Proto-MaxUp (PM) framework to minimize the issue of overfitting from the perspective of data augmentation and a feature extraction backbone for tree species classification. Taking Gaofeng Forest Farm (GFF) in Nanning City, Guangxi Province, as the study area, nine tree species, cutting site, and road were classified. First, by analyzing the effects of a series of popular data augmentation methods and their combinations in different parts of the P-Net, several effective data augmentation pools were established. Then, the pools aforementioned were combined with PM to obtain the best classification performance. To verify the robustness and validity of the proposed strategy, we applied PM to the other four popular public hyperspectral datasets and achieved excellent results. Finally, this efficient data augmentation method was used in different feature extraction backbones. The results show that the classification accuracy was greatly improved with the optimal backbone (overall accuracy (OA) and Kappa, are 98.08% and 0.9789, respectively), and the difference between training accuracy and test accuracy is less than 2%. It is concluded that the accurate and fine classification for multiple tree species can be realized by the PM data augmentation strategy and backbone proposed in this article.